deploy (#365 )

Reviewed-on: #365
deployment to prod (#357 )
2026-09-29 13:06:27 +02:00 · 2026-09-29 12:01:56 +02:00 · 2026-09-29 11:54:06 +02:00 · 2026-09-29 11:49:18 +02:00 · 2026-09-29 02:48:03 -07:00 · 2026-09-29 11:04:58 +02:00
73 changed files with 7952 additions and 558 deletions
@@ -88,7 +88,9 @@ RUN CGO_ENABLED=1 make build VERSION="$VERSION" GO_LDFLAGS='-extldflags "-static
 # alpine:3.21, 2026-03-17
 FROM alpine:3.21@sha256:c3f8e73fdb79deaebaa2037150150191b9dcbfba68b4a46d70103204c53f4709

-RUN apk --no-cache add ca-certificates
+# su-exec 0.2-r3 (Alpine 3.21), 2026-09-29: the entrypoint runs the app
+# as webhooker with it.
+RUN apk --no-cache add ca-certificates su-exec=0.2-r3

 # Create non-root user
 RUN addgroup -g 1000 -S webhooker && \
@@ -99,17 +101,34 @@ WORKDIR /app
 # Copy binary from builder
 COPY --from=builder /build/bin/webhooker /app/webhooker

+# Not under /app, which belongs to webhooker: this script runs as root.
+COPY deploy/docker-entrypoint.sh /usr/local/bin/docker-entrypoint.sh
+
 # Create data directory for all SQLite databases (main app DB +
 # per-webhook event DBs). DATA_DIR defaults to /var/lib/webhooker.
 RUN mkdir -p /var/lib/webhooker

 RUN chown -R webhooker:webhooker /app /var/lib/webhooker

-USER webhooker
+# No USER: the entrypoint starts as root to make the data directory
+# webhooker's, then runs the app as webhooker.

 EXPOSE 8080

+# The binary defaults BIND_ADDRESS to 127.0.0.1, which is right for a
+# bare host: the cleartext listener serves the admin UI and the
+# unauthenticated receiver, so it must not appear on every interface
+# of a machine that configured nothing. A container is the other case.
+# Its network namespace is already the isolation boundary, so binding
+# every address inside it exposes nothing; what decides exposure is
+# the publish flag, and `-p 127.0.0.1:8080:8080` is the operator's
+# control there. Shipping the image on loopback would buy no security
+# and would make the process unreachable through its own published
+# port.
+ENV BIND_ADDRESS=0.0.0.0
+
 HEALTHCHECK --interval=30s --timeout=3s --start-period=5s --retries=3 \
    CMD wget --no-verbose --tries=1 --spider http://localhost:8080/.well-known/healthcheck || exit 1

+ENTRYPOINT ["/usr/local/bin/docker-entrypoint.sh"]
 CMD ["/app/webhooker"]
@@ -7,6 +7,13 @@ services, durably stores them, and delivers them to configured targets
 with retry support, logging, and observability. Category: infrastructure
 / web service. License: MIT.

+Each entrypoint is a version 4 UUID served at `/webhook/{uuid}`, and
+that UUID is the entrypoint's only credential. webhooker does not use
+shared secrets, HMAC signatures or token headers on the receiver, and
+will not add them — read
+[The entrypoint URL is the authentication secret](#the-entrypoint-url-is-the-authentication-secret)
+before deploying one.
+
 ## Getting Started

 ### Prerequisites
@@ -37,9 +44,9 @@ make bootstrap
 # Run all checks (test, lint, format check)
 make check

-# Run in development mode. DATA_DIR defaults to /var/lib/webhooker in
-# every environment, so set it (in .env or the shell) to a writable
-# directory when running from a clone.
+# Run the server from the clone. DATA_DIR defaults to
+# /var/lib/webhooker in every environment, so set it (in .env or the
+# shell) to a writable directory.
 DATA_DIR=./data make dev

 # Build Docker image
@@ -71,11 +78,22 @@ make clean      # Remove bin/
 ### Configuration

 All configuration is via environment variables. For local development,
-you can place variables in a `.env` file in the project root (loaded
-automatically via `godotenv/autoload`).
+you can place variables in a `.env` file in the process working
+directory, read once at startup before anything else looks at the
+environment.
+
+The file is optional and having none is the normal case for a
+deployment. A file that is there but cannot be parsed aborts startup
+with a message naming it, because a single malformed line makes none
+of the file apply: every variable in it silently reverts to its
+default, which is exactly the failure [Invalid values abort
+startup](#invalid-values-abort-startup) exists to prevent, for all of
+them at once. A variable already present in the real environment wins
+over the file's value for the same name.

 The environment is selected by setting `WEBHOOKER_ENVIRONMENT` to `dev`
-or `prod` (default: `dev`). The setting controls exactly one behavior:
+or `prod` (default: `prod`; `dev` must be set explicitly). The setting
+controls exactly one behavior:

 | Behavior | `dev`                   | `prod`           |
 | -------- | ----------------------- | ---------------- |
@@ -117,14 +135,15 @@ TTY detection, and security headers are always applied.

 | Variable                | Description                         | Default  |
 | ----------------------- | ----------------------------------- | -------- |
-| `WEBHOOKER_ENVIRONMENT` | `dev` or `prod`                     | `dev`    |
+| `WEBHOOKER_ENVIRONMENT` | `dev` or `prod`                     | `prod`   |
 | `PORT`                  | HTTP listen port                    | `8080`   |
+| `BIND_ADDRESS`          | IP address the HTTP listener binds. Loopback by default, so the cleartext listener is not published on every interface. The Docker image ships `0.0.0.0` instead. See [Bind address](#bind-address) | `127.0.0.1` (image: `0.0.0.0`) |
 | `DATA_DIR`              | Directory for all SQLite databases  | `/var/lib/webhooker` |
 | `DEBUG`                 | Enable debug logging                | `false`  |
 | `MAINTENANCE_MODE`      | Report `maintenanceMode: true` in the healthcheck JSON. It does not change how any request is served — no maintenance page exists | `false`  |
 | `METRICS_USERNAME`      | Basic auth username for `/metrics`. Must be set together with `METRICS_PASSWORD`; one without the other fails startup | `""`     |
 | `METRICS_PASSWORD`      | Basic auth password for `/metrics`. Must be set together with `METRICS_USERNAME`; one without the other fails startup | `""`     |
-| `SENTRY_DSN`            | Sentry error reporting DSN          | `""`     |
+| `SENTRY_DSN`            | Sentry error reporting DSN. Unset leaves error reporting off; a value the Sentry SDK cannot parse fails startup rather than serving with reporting silently off | `""`     |
 | `RETENTION_SWEEP_INTERVAL` | How often the retention reaper and archive sweeper run (Go duration, must be positive) | `1h` |
 | `SESSION_IDLE_TIMEOUT`  | Idle session timeout (Go duration)  | `24h`    |
 | `RECEIVER_RATE_LIMIT`   | Receiver requests/minute per IP per entrypoint (10x that per IP across the route) | `120` |
@@ -138,6 +157,11 @@ private and reserved ranges — RFC 1918, loopback, CGNAT, link-local and
 the rest — are refused, which stops a target from being used to make
 webhooker probe the network it sits in.

+Besides the private and reserved ranges, the default blocklist refuses
+public cloud metadata addresses: currently only `168.63.129.16`, Azure's
+WireServer, which serves an Azure VM its credentials. Because it is a
+public address, listing it in `ALLOWED_EGRESS_CIDRS` reopens it.
+
 That default is also inconvenient for the thing webhooker is mostly
 for: taking a public webhook and forwarding it to something on your own
 network. A container on the same Docker network, a box on `10.x`, a
@@ -176,15 +200,16 @@ Two things this setting cannot do:
  the list is always an allowlist; an empty list (the default) means
  every private and reserved range stays refused. Note that
  `0.0.0.0/0` gets you most of the way there anyway, per above.
- **It cannot open link-local, or a cloud metadata endpoint that
-  discloses credentials or user data.** An address is on the list below
-  when both of these hold: the provider fixes it, so it cannot collide
-  with anything you run; and reaching it hands out credentials, user
-  data or bootstrap material. Those stay blocked no matter what you
-  list, including when you list them outright or list a supernet such
-  as `0.0.0.0/0`, `::/0`, `fd00::/8` or `100.64.0.0/10`. Treat this as
-  best effort rather than a guarantee — it is a hand-maintained list
-  and the caveat below the table applies:
+- **It cannot open link-local, or a cloud metadata endpoint at a
+  non-public address that discloses credentials or user data.** An
+  address is on the list below when it is not a public address and both
+  of these hold: the provider fixes it, so it cannot collide with
+  anything you run; and reaching it hands out credentials, user data or
+  bootstrap material. Those stay blocked no matter what you list,
+  including when you list them outright or list a supernet such as
+  `0.0.0.0/0`, `::/0`, `fd00::/8` or `100.64.0.0/10`. Treat this as best
+  effort rather than a guarantee — it is a hand-maintained list and the
+  caveat below the table applies:

  | Blocked unconditionally | What it is |
  | ----------------------- | ---------- |
@@ -223,7 +248,8 @@ Two things this setting cannot do:
  encodings, which the default blocklist does not match. A publicly
  routable metadata address is not listed here, because nothing on this
  list can be reopened and blocking one that way would leave you no
-  escape hatch at all.
+  escape hatch at all; Azure's `168.63.129.16` is refused by the default
+  blocklist instead, as described above.

  This list is not exhaustive of every cloud's metadata address — if
  yours is not here, do not allowlist the block that contains it.
@@ -240,6 +266,68 @@ A set but unparseable value aborts startup. When the list is non-empty
 webhooker logs it at startup, blocks and all, so the hole is visible in
 the log of any deployment that has one.

+#### Bind address
+
+`BIND_ADDRESS` is the IP address the HTTP listener binds. The binary
+defaults to `127.0.0.1`, so a bare webhooker is reachable only from the
+host it runs on. The Docker image ships `ENV BIND_ADDRESS=0.0.0.0`
+instead — see below for why the two differ.
+
+That listener speaks **cleartext**, and it serves both the admin UI and
+the unauthenticated webhook receiver. webhooker terminates no TLS
+itself; a production deployment puts a reverse proxy in front of it
+(see
+[Deployment behind a reverse proxy](#deployment-behind-a-reverse-proxy)),
+and the proxy reaches it over loopback. A default that bound every
+interface would leave that cleartext port answering the internet
+alongside the proxy — the admin login form and the receiver, in the
+clear, on a port nobody chose to publish. Reaching webhooker from
+another host is therefore something you configure, not something you
+get by default.
+
+**In a container the answer is `0.0.0.0`, which is why the image ships
+that.** A container's network namespace is already the boundary the
+loopback default is reaching for: nothing outside the container gets to
+`0.0.0.0:8080` because of the namespace, whatever the process bound.
+Exposure is decided at the publish flag instead — `-p
+127.0.0.1:8080:8080` rather than `-p 8080:8080` — which is the
+operator's to choose and is what
+[Running with Docker](#running-with-docker) shows. A loopback bind
+inside a container buys nothing and makes the process unreachable
+through its own published port.
+
+The value must be an IP address literal:
+
+- `127.0.0.1` — loopback only (the binary's default). Use this with a
+  reverse proxy on the same host.
+- `0.0.0.0` — every IPv4 address. The image's default; on a bare host,
+  only behind a firewall on the port.
+- `::` — every address, IPv6 and (on Linux, with the default
+  `net.ipv6.bindv6only=0`) IPv4 as well.
+- A specific address such as `10.0.0.5` — that interface only.
+
+An **empty** value is treated as unset, as everywhere else here, and
+takes the default. In a container that matters: `BIND_ADDRESS=` throws
+away the image's `0.0.0.0` and falls back to the binary's
+`127.0.0.1`, which is the one quiet failure this setting has — see
+[Running with Docker](#running-with-docker).
+
+Hostnames are **not** accepted. `localhost` aborts startup rather than
+being resolved: which of `127.0.0.1` and `::1` it means differs by
+host, a name can resolve to several addresses of which only one could
+be bound, and the answer can change under a running process. A value
+carrying a port (`127.0.0.1:8080`) is likewise rejected — the port is
+`PORT`'s business. Any unparseable value aborts startup; see
+[Invalid values abort startup](#invalid-values-abort-startup).
+
+An address that parses but is not assigned to this host — say
+`10.0.0.5` on a machine that has no such interface — is a valid
+literal, so it reaches the listener and fails there. The process logs
+the bind error and exits non-zero rather than staying up with nothing
+listening. The effective value is in the `bindAddress` field of the
+startup log line, which is the way to check what a running deployment
+actually bound.
+
 #### Metrics credentials

 `METRICS_USERNAME` and `METRICS_PASSWORD` are set together or not at
@@ -319,10 +407,9 @@ the bucket is. See [Rate Limiting](#rate-limiting).

 The remedy is to set `TRUSTED_PROXIES` to your reverse proxy's
 address, which restores per-client buckets. webhooker logs a warning
-at startup whenever `TRUSTED_PROXIES` is empty, in every environment —
-not only when `WEBHOOKER_ENVIRONMENT=prod`, because that variable
-defaults to `dev` and an operator who never set it is precisely the
-one at risk. The warning is informational when nothing proxies to the
+at startup whenever `TRUSTED_PROXIES` is empty, in every environment,
+because behind a proxy every client shares one bucket in `dev` and
+`prod` alike. The warning is informational when nothing proxies to the
 process: with no proxy in front, the peer address is the client's own
 and the buckets are already per-client. See
 [Rate Limiting](#rate-limiting) for what each limit shares.
@@ -397,11 +484,24 @@ additionally be a number in the range 1–65535,
 `RECEIVER_RATE_LIMIT` must be at least 1,
 `RETENTION_SWEEP_INTERVAL` must be greater than zero (it is a ticker
 period, so `0s` or a negative value would crash the reaper after
-startup), and every entry in `TRUSTED_PROXIES` and
-`ALLOWED_EGRESS_CIDRS` must be a CIDR block or a bare IP address.
+startup), every entry in `TRUSTED_PROXIES` and
+`ALLOWED_EGRESS_CIDRS` must be a CIDR block or a bare IP address, and
+`BIND_ADDRESS` must be an IP address literal — `localhost`,
+`127.0.0.1:8080` and `10.0.0.0/8` are each rejected rather than
+resolved, split, or narrowed to something they do not say — and
+`SENTRY_DSN` must parse as a Sentry DSN.
 `SESSION_IDLE_TIMEOUT` is the exception: a
 non-positive value there means idle expiry is disabled, not invalid.

+`SENTRY_DSN` is checked with the Sentry SDK's own parser, the same call
+the SDK makes on the DSN it is later handed, so what configuration
+accepts is exactly what will initialise. A typo in it is the one
+configuration mistake nothing downstream can ever notice — the variable
+is still set, so every later signal reports error reporting as on while
+no report is being sent — which is why it aborts rather than starting
+with reporting off. Leaving it unset is not a mistake and not affected:
+error reporting is simply off and startup is normal.
+
 Boolean variables (`DEBUG`, `MAINTENANCE_MODE`) accept exactly the
 spellings Go's `strconv.ParseBool` accepts — `1`, `t`, `T`, `TRUE`,
 `true`, `True`, `0`, `f`, `F`, `FALSE`, `false`, `False` — and nothing
@@ -438,6 +538,12 @@ its Argon2id hash. There is no second account and no forgot-password
 flow, so the banner and the reset command below are the only two ways
 in.

+A start that finds no `webhooker.db` in `DATA_DIR` also logs
+`created a new, empty database` at `WARN`, with the file's path,
+shortly before the banner. On a deployment that has run before, that
+line means `DATA_DIR` was empty, most often because its volume is not
+mounted.
+
 #### Recovering a lost admin password

 `webhooker resetpw` sets an existing account's password from the
@@ -452,8 +558,9 @@ printf '%s' "$NEW_PASSWORD" | \
    DATA_DIR=/var/lib/webhooker webhooker resetpw admin
 ```

-In a container it is the same binary, which the image sets as `CMD`
-rather than `ENTRYPOINT`, so the whole command has to be given:
+In a container it is the same binary. The image's `CMD` is
+`/app/webhooker`, and a command given to `docker run` replaces all of
+it, so the whole command has to be given:

 ```bash
 docker run --rm -v webhooker-data:/var/lib/webhooker \
@@ -543,20 +650,274 @@ decision:

 ```bash
 docker run -d \
-  -p 8080:8080 \
+  -p 127.0.0.1:8080:8080 \
  -v /path/to/data:/var/lib/webhooker \
-  -e WEBHOOKER_ENVIRONMENT=prod \
+  -e BIND_ADDRESS=0.0.0.0 \
  webhooker:latest
 ```

-The container runs as a non-root user (`webhooker`, UID 1000), exposes
-port 8080, and includes a health check against
-`/.well-known/healthcheck`. The `/var/lib/webhooker` volume holds all
-SQLite databases: the main application database (`webhooker.db`), the
-per-webhook event databases (`events-{uuid}.db`), and any archive
-databases written by `database` targets (`archive-{uuid}.db`). Mount
-this as a persistent volume to preserve data across container
-restarts.
+**The image and the bare binary default `BIND_ADDRESS` differently, on
+purpose.** The binary defaults to `127.0.0.1`; the image ships
+`ENV BIND_ADDRESS=0.0.0.0`, so the `-e BIND_ADDRESS=0.0.0.0` above is
+belt-and-braces and the command works without it.
+
+The two cases are not the same question. On a bare host, `0.0.0.0`
+puts the cleartext admin UI and the unauthenticated receiver on every
+interface of the machine, which is what the loopback default exists to
+prevent. In a container, the network namespace is already that
+boundary: nothing outside reaches `0.0.0.0:8080` because of the
+namespace, not because of the bind. What decides exposure there is the
+**publish flag**, and that is the line to get right.
+
+So publish to `127.0.0.1:8080` rather than `8080`. A bare
+`-p 8080:8080` opens the port on every interface of the host — through
+firewall rules too, since Docker's forwarding rules are inserted ahead
+of most host firewalls. Publish to the host address your reverse proxy
+connects from, and nothing wider.
+
+**An empty `BIND_ADDRESS` is treated as unset**, like every other
+variable here, so `-e BIND_ADDRESS=` does not mean "keep the image
+default" — it discards the image's `0.0.0.0` and falls back to the
+_binary's_ `127.0.0.1`. In a container that is the failure below, and
+nothing in the logs names the variable. A templated Compose file or a
+`.env` line with an empty value is the usual way in. Either set a
+literal or leave the variable out entirely.
+
+Overriding `BIND_ADDRESS` to a loopback address in a container — by
+that route or deliberately — makes the container unreachable from
+outside its namespace even with `-p`. The published port answers
+nothing, and the health check fails too: it requests
+`http://localhost:8080`, `localhost` resolves to `::1` first, and a
+`127.0.0.1` bind is not listening there. The container then goes
+`unhealthy` about **65 seconds** after start — from `HEALTHCHECK
+--start-period=5s --interval=30s --retries=3`, so failing probes at
+5s, 35s and 65s, and `unhealthy` on the third. (Docker's probe cadence
+during the start period has changed between versions; re-derive from
+those three values rather than trusting the figure. Measured at 65s on
+Docker 29.7.2.) A container `unhealthy` with `connection refused` in
+its health log, or a published port that resets connections, is this.
+
+The app runs as a non-root user (`webhooker`, UID 1000), exposes port
+8080, and includes a health check against `/.well-known/healthcheck`.
+The `/var/lib/webhooker` volume holds all SQLite databases: the main
+application database (`webhooker.db`), the per-webhook event databases
+(`events-{uuid}.db`), and any archive databases written by `database`
+targets (`archive-{uuid}.db`). Mount this as a persistent volume to
+preserve data across container restarts.
+
+**The container sets its data directory's owner and mode itself
+before the app starts**, so a host directory can be mounted as it is,
+whoever owns it. The image's `ENTRYPOINT`,
+`deploy/docker-entrypoint.sh`, starts as root, creates `DATA_DIR` if
+it is missing, gives the directory and anything in it that belongs to
+another user to `webhooker`, sets the directory to `0750`, and only
+then runs the app as `webhooker`. Started with `--user`, it changes
+nothing and runs the app as that user.
+
+**The file modes are not yours to set, and do not depend on the
+directory.** `webhooker.db` holds target configuration in plaintext —
+bearer tokens, API keys, Slack webhook URLs — along with the session
+encryption key, so webhooker creates every SQLite file it owns `0600`:
+each database and both of its `-wal` and `-shm` sidecars, across all
+three tiers. Files an earlier build left `0644` are tightened when
+they are opened. The directory's `0750` is defence in depth — it stops
+other local users listing the directory and learning your webhook
+UUIDs from the `events-{uuid}.db` filenames — not the barrier
+protecting the credentials.
+
+### Running under upaas
+
+[upaas](https://git.eeqj.de/sneak/upaas) builds the image from this
+repository's `Dockerfile` and runs it. The app needs:
+
+- **Network and port:** add no port mapping in upaas. upaas publishes
+  every mapped port on all interfaces of the host
+  ([upaas issue 113](https://git.eeqj.de/sneak/upaas/issues/113)),
+  which would put the plain-HTTP admin UI and receiver there. Instead,
+  set the app's Docker Network in upaas to your reverse proxy's Docker
+  network; the proxy then reaches the app at `upaas-` followed by the
+  app name, port `8080`. Leave `PORT` unset: the image's health check
+  probes `8080`.
+- **Volume:** one host directory mounted at `/var/lib/webhooker`.
+- **Environment variables:**
+    - `WEBHOOKER_ENVIRONMENT=prod`
+    - `TRUSTED_PROXIES`: your reverse proxy's address on that Docker
+      network. The `remoteIP` field of the `http request` log line for a
+      request that came through the proxy shows it; the health check's
+      own lines show `::1`. See [Trusted proxies](#trusted-proxies).
+    - Leave `BIND_ADDRESS` and `DATA_DIR` unset: the image sets
+      `BIND_ADDRESS` to `0.0.0.0`, and `DATA_DIR` defaults to
+      `/var/lib/webhooker`.
+    - Everything else is optional; see [Configuration](#configuration).
+- **Health check:** the image's own, which requests
+  `/.well-known/healthcheck`. upaas reads the container's health 60
+  seconds after a deploy and marks the deploy failed unless it is
+  `healthy`.
+- **First run:** the first start prints the `admin` password once, in
+  the banner described under [The admin account](#the-admin-account),
+  to the container's log. upaas names the container `upaas-` followed
+  by the app name, so for an app named `webhooker`:
+
+    ```bash
+    docker logs upaas-webhooker
+    ```
+
+  If the password is lost, stop the container, set a new password with
+  the app's own image and volume, and start it again (see
+  [Recovering a lost admin password](#recovering-a-lost-admin-password)):
+
+    ```bash
+    docker stop upaas-webhooker
+    docker run --rm --volumes-from upaas-webhooker \
+        "$(docker inspect -f '{{.Image}}' upaas-webhooker)" \
+        /app/webhooker resetpw -generate admin
+    docker start upaas-webhooker
+    ```
+
+## Deployment behind a reverse proxy
+
+webhooker terminates no TLS of its own. It serves plaintext HTTP and
+expects a reverse proxy in front of it, which is the deployment it is
+built for: the proxy holds the certificate, and webhooker binds
+loopback where only the proxy can reach it.
+
+Five things have to be right. Each one is silent when it is wrong —
+the service comes up, serves pages, and is broken in a way nothing
+reports.
+
+1. **Bind or firewall the app port.** The binary binds `127.0.0.1` by
+   default, so the cleartext listener is not published beside the
+   proxy. The image binds `0.0.0.0` inside its own network namespace
+   and relies on the publish address instead —
+   `-p 127.0.0.1:8080:8080`. Either way the port must reach the proxy
+   and nothing else; widen it only with a firewall or a publish
+   address in front of it. A cleartext port answering the internet
+   serves the admin login form and the unauthenticated receiver with
+   no TLS at all, and the proxy in front of it changes nothing about
+   that.
+2. **Make sure the environment is not `dev` (leave
+   `WEBHOOKER_ENVIRONMENT` unset or set it to `prod`), and make sure
+   the proxy sends `X-Forwarded-Proto`.** These are two requirements,
+   not one. The environment setting decides CORS and nothing else:
+   `dev` answers every origin with `Access-Control-Allow-Origin: *`
+   (without credentials), which a server-rendered production
+   deployment has no use for, and `prod` — the default — disables it.
+   Cookie `Secure` and the strict Origin/Referer mode are **not** tied
+   to it — they are decided per request from the transport, which
+   behind a proxy means the `X-Forwarded-Proto` header. The block below
+   sets it; without it every request is read as plaintext and cookies
+   ship without `Secure`. See [Configuration](#configuration).
+3. **Set `TRUSTED_PROXIES` to the proxy's address.** Unset, every rate
+   limiter keys on the connecting peer, which behind a proxy is the
+   proxy on every request: all clients collapse into one global bucket
+   per limit and the receiver's per-IP limits become service-wide
+   ceilings. See [Trusted proxies](#trusted-proxies). List the proxy
+   and nothing else.
+4. **Send `Host` as `$http_host`, not `$host`.** `$host` strips the
+   port. webhooker's Origin/Referer check compares against the host it
+   was given, so on any port other than 443 `$host` makes every form
+   POST — including login — fail with `403 origin invalid`, with
+   nothing in the error naming the cause.
+5. **Keep the proxy's access log.** webhooker's own access log records
+   the peer address, which behind a proxy is always the proxy. The
+   proxy's log is the only record of which client sent what. nginx's
+   default `combined` format already logs `$remote_addr`; do not
+   replace it with one that drops the client address, and retain those
+   logs as long as you would want to answer a question about traffic.
+
+### nginx
+
+Complete server block. Replace the `server_name` and the two
+certificate paths.
+
+```nginx
+server {
+    listen 443 ssl;
+    listen [::]:443 ssl;
+    http2 on;  # nginx 1.25.1+; older: listen 443 ssl http2;
+
+    server_name webhooker.example.com;
+
+    ssl_certificate     /etc/ssl/certs/webhooker.example.com.crt;
+    ssl_certificate_key /etc/ssl/private/webhooker.example.com.key;
+    ssl_protocols       TLSv1.2 TLSv1.3;
+
+    # webhooker caps form POST bodies at 1 MB. nginx's default happens
+    # to match, so leaving this out breaks nothing today — but if you
+    # ever raise webhooker's cap, this is the limit you will still be
+    # hitting, and the rejection is nginx's HTML page rather than
+    # webhooker's message.
+    client_max_body_size 1m;
+
+    # $remote_addr is the client. webhooker's own log records this
+    # proxy and nothing else, so this file is the only place the
+    # client's address is written down.
+    access_log /var/log/nginx/webhooker.access.log combined;
+
+    location / {
+        # A literal address, not localhost: with BIND_ADDRESS at its
+        # 127.0.0.1 default, a localhost that resolves to ::1 first
+        # gets connection refused.
+        proxy_pass http://127.0.0.1:8080;
+
+        # $http_host, NOT $host. $host drops the port and every form
+        # POST fails with 403 origin invalid on any port but 443.
+        proxy_set_header Host              $http_host;
+        proxy_set_header X-Forwarded-For   $proxy_add_x_forwarded_for;
+        proxy_set_header X-Forwarded-Proto $scheme;
+
+        # Above webhooker's own 60s request timeout, so its 503
+        # reaches the client instead of nginx cutting the connection
+        # first and answering 504.
+        proxy_read_timeout 70s;
+    }
+}
+
+server {
+    listen 80;
+    listen [::]:80;
+    server_name webhooker.example.com;
+    return 308 https://$host$request_uri;
+}
+```
+
+`X-Forwarded-For` must be **appended** to, which
+`$proxy_add_x_forwarded_for` does. webhooker reads no other forwarded
+client header: `X-Real-IP` and `True-Client-IP` are ignored from every
+peer, so setting them has no effect. See
+[Trusted proxies](#trusted-proxies) for how the chain is walked.
+
+`X-Forwarded-Proto: https` is what tells webhooker the request arrived
+over TLS, which decides the `Secure` flag on both the session and CSRF
+cookies and the strict Origin/Referer mode. Without it, requests are
+treated as plaintext and the cookies ship without `Secure`. Unlike
+`X-Forwarded-For`, this header is read from any peer and is not gated
+by `TRUSTED_PROXIES`, so the proxy must overwrite whatever a client
+sent — `$scheme` above does.
+
+With that block, webhooker's environment is:
+
+```sh
+BIND_ADDRESS=127.0.0.1   # the default; stated here to be explicit
+TRUSTED_PROXIES=127.0.0.1
+```
+
+If nginx runs on another host, `BIND_ADDRESS` becomes the address it
+connects to, `TRUSTED_PROXIES` becomes nginx's address, and the port
+must be firewalled to that address — the traffic between them is
+cleartext.
+
+### HSTS is always sent, and is not configurable
+
+Every response carries
+`Strict-Transport-Security: max-age=63072000; includeSubDomains; preload`.
+Two years, every subdomain, and a `preload` token. There is no setting
+that changes or suppresses it.
+
+This is worth knowing before the first request reaches a browser: a
+client that sees it once will refuse plaintext HTTP to that hostname —
+and to every subdomain of it — for two years, whatever else is served
+there. Terminate TLS on a hostname you are prepared to keep on HTTPS.

 ## Backup, Restore, and Upgrades

@@ -581,9 +942,20 @@ is both the simplest and the only complete rule:
 `events-3f2a1c9e-....db`. The only other file is `webhooker.lock`, the
 always-empty [single-instance lock](#single-instance-lock); it holds no
 state and is not part of the backup set — a copied one is stale and
-blocks nothing. No `-wal` or `-shm` files are produced (see below); a
-transient `{name}.db-journal` may exist beside a database while a write
-is in flight and is not part of the backup set either.
+blocks nothing.
+
+**`-wal` and `-shm` sidecars.** Every database runs in WAL journal mode,
+so while the service is running each `{name}.db` has a `{name}.db-wal`
+and a `{name}.db-shm` beside it. **`-wal` is part of the database, not a
+scratch file**: it holds committed transactions that are not yet in the
+`.db`, so a copy of the `.db` without its `-wal` is missing data and may
+have no readable schema at all. `-shm` is regenerable, but there is no
+reason to separate the two — copy the directory and you have them.
+
+A clean shutdown closes every database, which checkpoints and removes
+its sidecars; a killed or crashed instance leaves them, and they must be
+carried with the `.db`. An archive the service has not opened since a
+crash keeps that crash's sidecars, even across a later clean stop.

 Configuration is **not** in `DATA_DIR` — it comes from the environment
 and from a `.env` file read out of the process working directory. Back
@@ -591,17 +963,19 @@ that up with your deployment config, separately.

 ### A hot copy is not safe

-No `journal_mode` pragma is ever issued on any database webhooker opens,
-so all of them run on SQLite's default rollback journal. There is no
-WAL. The main and event databases are also held open for the entire
-process lifetime — `WebhookDBManager` caches event database handles and
-closes them only on webhook deletion or shutdown — so "it looked idle"
-is not a guarantee that nothing was mid-transaction.
+Every database webhooker opens runs in WAL journal mode. The main and
+event databases are also held open for the entire process lifetime —
+`WebhookDBManager` caches event database handles and closes them only on
+webhook deletion or shutdown — so "it looked idle" is not a guarantee
+that nothing was mid-transaction.

 That means `cp`, `rsync`, `tar` or a filesystem snapshot taken against a
-running instance can capture a database mid-transaction and yield a file
-that is corrupt or missing state the journal would have rolled back. Use
-one of the two procedures below instead.
+running instance can capture a database and its `-wal` at two different
+instants and yield a file that is corrupt or missing state. Copying a
+`.db` on its own is worse and fails loudly: recently written pages,
+including the schema itself on a young database, live in the `-wal`, so
+the copy reads back as an empty or table-less database. Use one of the
+two procedures below instead.

 **Stop, copy, start.** The simplest, needs no extra tooling, and the
 only one that gives a single point in time across every file:
@@ -620,24 +994,46 @@ for db in /path/to/data/*.db; do
 done
 ```

-`.backup` takes the proper locks and writes a consistent file. Two
-caveats. First, the runtime image is `alpine:3.21` with only
-`ca-certificates` added — the `sqlite3` CLI is **not** in it, so run
-this on the host against the volume path, or from a throwaway container
-that mounts the volume. Second, each file is captured at its own
-instant, so a webhook created or an event delivered between two files
-being copied lands in one and not the other. If you need the whole set
-coherent as of a single moment, stop the service.
+`.backup` reads through the WAL and writes a single consistent file with
+no sidecars of its own, so the destination is complete as it stands.
+Two caveats. First, the runtime image is `alpine:3.21` with only
+`ca-certificates` and `su-exec` added — the `sqlite3` CLI is **not** in
+it, so run this on the host against the volume path, or from a
+throwaway container that mounts the volume. Second, each file is
+captured at its own instant, so a webhook created or an event delivered
+between two files being copied lands in one and not the other. If you
+need the whole set coherent as of a single moment, stop the service.
+
+Note that `sqlite3 <db> .dump` is **not** one of these procedures: it is
+an export, it holds a read transaction open for as long as it runs, and
+it pins the WAL against checkpointing for that whole time. It is safe to
+run — it does not block ingestion — but back up with `.backup` or a
+stopped copy.

 Archive databases are the one exception the service is built for: the
-archive writer closes its handle after each write (debounced to at most
-one reopen per second), so an operator can move `archive-{uuid}.db`
-away for offline retention while the service runs, and it is recreated
-on the next write (see
-[Database Architecture](#database-architecture)). That is a
+archive writer closes and reopens its handle around writes (debounced
+to at most one reopen per second), so an operator can move
+`archive-{uuid}.db` away for offline retention while the service runs,
+and it is recreated on the next write. See
+[Database Architecture](#database-architecture). That is a
 move-the-file-away workflow, not a substitute for the backup procedures
 above.

+**Move the sidecars with it.** Under WAL that workflow is no longer a
+single file, and the common case is the dangerous one. The reopen
+happens on the *next* write after the debounce window elapses, so after
+the last write of a burst nothing checkpoints: measured, 20 s after ten
+events the `archive-….db` was 4096 bytes — a header, no table — with
+all ten rows sitting in a 189 KB `-wal`. Copying the `.db` alone at that
+moment yields a file that opens with `no such table: archived_events`.
+The file becomes self-contained again when the handle closes, which
+happens on the next write past the debounce window, when the connection
+pool retires the idle connection (about a minute after the last write),
+or at the idle archive sweep — measured, the same file was a complete
+20 KB `.db` with no sidecars about a minute after its last write. A
+clean stop closes it too. So either move `archive-{uuid}.db` together
+with any `-wal`/`-shm` beside it, or wait until there are none.
+
 ### Restore

 1. Stop the service.
@@ -651,24 +1047,20 @@ above.
   restored without `webhooker.db` are simply orphaned; nothing
   references their UUIDs.

-3. Do not carry `*.db-journal` files into the restore. Backups taken by
-   either procedure above are self-consistent and do not need one.
+3. Carry any `*.db-wal` and `*.db-shm` files that are in the backup.
+   They are part of the database, and dropping a `-wal` silently
+   discards every transaction it still holds. An `.backup` set will not
+   contain any: it writes a single consolidated file per database. A
+   stop-and-copy set normally has none, because a clean stop closes
+   every database and checkpoints its sidecars away; the exception is an
+   archive not opened since a crash. A copy salvaged from a crashed
+   instance has them for everything, and needs all of them.

-4. **Fix ownership.** The container runs as the non-root `webhooker`
-   user, UID 1000 / GID 1000. Restored files must be owned by (or
-   writable by) that UID, and so must the directory itself — SQLite
-   creates the rollback journal beside the database, so a writable file
-   inside a directory it cannot write is not enough:
-
-    ```bash
-    chown -R 1000:1000 /path/to/data
-    ```
-
-    Restoring as `root` on the host and forgetting this step is the
-    usual way a restore fails.
-
-5. Start the service. `AutoMigrate` runs against each restored database
-   as it is opened.
+4. Start the service. The container gives the directory and the
+   restored files to the `webhooker` user before the app starts,
+   whoever restored them (see
+   [Running with Docker](#running-with-docker)). `AutoMigrate` runs
+   against each restored database as it is opened.

 ### Upgrades

@@ -696,6 +1088,18 @@ Upgrade procedure:
   `curl -s http://host:8080/.well-known/healthcheck` reports the
   version it was stamped with (see [Version stamping](#version-stamping)).

+**Upgrading past the introduction of `BIND_ADDRESS`:** earlier versions
+always bound every interface. **Container deployments are unaffected**
+— the image ships `ENV BIND_ADDRESS=0.0.0.0`, so a `docker run` or
+Compose service that worked before still works with nothing changed.
+
+A **bare binary** is the case that changes: the listener now binds
+`127.0.0.1` unless `BIND_ADDRESS` says otherwise, so a deployment that
+relied on reaching it from another host becomes unreachable until it
+sets the address the proxy connects to. Check the `bindAddress` field
+of the startup log to see what a running process bound. See
+[Bind address](#bind-address).
+
 **Downgrade is unsupported.** Once a newer binary has migrated the files
 there is no way to move them back. `AutoMigrate` is additive — it adds
 tables, columns and indexes and never drops or rewrites them — so an
@@ -777,14 +1181,38 @@ backups at rest and restrict who can read them.

 ## The entrypoint URL is the authentication secret

-The receiver verifies nothing about an inbound request. The UUID in an
-entrypoint's URL is its credential: anyone who holds that URL can
-submit events to it, and the receiver checks nothing else about the
-sender. Treat an entrypoint URL the way you would treat an API token.
+**The entrypoint UUID is the credential, and it is the only one.**
+webhooker mints a version 4 UUID per entrypoint and serves it at
+`/webhook/{uuid}`. Possession of that URL is the authentication:
+anyone who holds it can submit events to the entrypoint, and the
+receiver verifies nothing else about the sender.

-There is no way to rotate the UUID in place. To retire one, delete the
-entrypoint (or deactivate it, which answers `410`) and create a new
-one, then point the sender at the new URL.
+There is no shared secret, no HMAC signature, no bearer token and no
+second factor on the receiver, and none will be added. This was
+considered and rejected; the implementation that existed was removed
+in [PR #279](https://git.eeqj.de/sneak/webhooker/pulls/279), closing
+[issue #67](https://git.eeqj.de/sneak/webhooker/issues/67) and
+[issue #241](https://git.eeqj.de/sneak/webhooker/issues/241). A
+proposal to reintroduce any of them — including as "defence in depth"
+alongside the UUID — is answered by this section. Inbound signature
+headers a sender sends anyway (`X-Hub-Signature` and its
+per-provider equivalents) are stored and forwarded as ordinary
+headers; nothing checks them.
+
+What that means for an operator:
+
+- **The URL is a capability, so treat it as a secret.** Keep it out of
+  logs, ticket bodies, chat messages and screenshots. Anyone who reads
+  it anywhere can post events as that sender.
+- **Rotating means minting a new entrypoint, not changing a key.**
+  There is no way to rotate the UUID in place. To retire one, delete
+  the entrypoint (or deactivate it, which answers `410`) and create a
+  new one, then point the sender at the new URL.
+- **A sender that cannot be given a secret URL is a constraint on that
+  integration, not a reason to change this.** If a service only
+  supports signed payloads to a well-known URL, raise it as its own
+  problem — pick a different integration path, or accept that it
+  cannot be used. It is not grounds to reintroduce shared secrets.

 ## Entrypoints

@@ -794,8 +1222,15 @@ standard: normalized scripts in `script/` are the entrypoints for the
 development workflow. Ten of the Makefile's seventeen targets are thin
 shims that call them; `build`, `run`, `dev`, `deps`, `clean`, `css` and
 `version` are inline commands with no script behind them, though
-`build` and `version` both take their value from `script/version`. We
-provide:
+`build` and `version` both take their value from `script/version`.
+
+`make check` needs the third-party browser assets in `static/`, which
+are not committed, so run `make bootstrap` (or just `make assets`) once
+after cloning. Without them the tests fail with a message naming that
+remedy. `make check` does not fetch them itself because it must not
+change any files in the repo.
+
+We provide:

 - `script/bootstrap` — install all dependencies (idempotent)
 - `script/setup` — make a fresh clone ready for development
@@ -862,6 +1297,16 @@ webhooker solves this by acting as a durable intermediary:
   backoff. Every delivery attempt is logged with status codes, response
   bodies, and timing.

+   **That guarantee is at-least-once, not exactly-once.** When a send
+   reaches its target but the write recording that outcome fails, the
+   delivery is deliberately left in a recoverable state rather than
+   marked done — losing a delivery is the worse failure — so the
+   pending sweep picks it up about fifteen minutes later, or the next
+   restart does, and the target receives a payload it already got.
+   webhooker adds no delivery identifier of its own to an outbound
+   request, so **make your receiver idempotent** against whatever the
+   payload itself carries.
+
 3. **Observability** — Full request/response logging for every webhook
   received and every delivery attempted. Prometheus metrics expose
   volume, latency, and error rates. The web UI provides real-time
@@ -1094,7 +1539,7 @@ events should be forwarded.
 | `type`           | TargetType | One of: `http`, `slack`, `database`, `log` |
 | `active`         | boolean    | Whether deliveries are enabled (default: true) |
 | `config`         | JSON text  | Type-specific configuration |
-| `max_retries`    | integer    | Maximum retry attempts for `http` and `slack` targets (0 = fire-and-forget, >0 = retries with backoff and a circuit breaker). Ignored by `database` and `log` targets |
+| `max_retries`    | integer    | Total delivery attempts for `http` and `slack` targets, not retries on top of the first: 0 is a single fire-and-forget attempt with no retries and no circuit breaker, and a value of N makes N attempts in all, with exponential backoff and a per-target circuit breaker. Ignored by `database` and `log` targets |
 | `max_queue_size` | integer    | Stored and shown on the target's detail view, but not enforced anywhere yet: nothing in the delivery engine consults it. Queue depth is set by the two fixed 10,000-entry channels |

 **Relations:** Belongs to Webhook. Has many Deliveries.
@@ -1102,12 +1547,12 @@ events should be forwarded.
 **Target types:**

 - **`http`** — Forward the event as an HTTP POST to a configured URL.
-  Behavior depends on `max_retries`: when `max_retries` is 0 (the
-  default), the target operates in fire-and-forget mode — a single
-  attempt with no retries and no circuit breaker. When `max_retries` is
-  greater than 0, failed deliveries are retried with exponential backoff
-  up to `max_retries` attempts, protected by a per-target circuit
-  breaker.
+  `max_retries` is the total number of delivery attempts, not retries on
+  top of the first: when `max_retries` is 0 (the default), the target
+  operates in fire-and-forget mode, a single attempt with no retries and
+  no circuit breaker; a value of N makes up to N attempts in all,
+  retrying failed deliveries with exponential backoff and protecting them
+  with a per-target circuit breaker.
 - **`slack`** — Post the event as a formatted message to a
  Slack-compatible incoming webhook URL (`webhookUrl` in `config`). It
  is built on the same HTTP core as `http` and honours `max_retries`
@@ -1287,6 +1732,28 @@ retries) is individually logged for full observability.

 **Relations:** Belongs to Delivery.

+#### Event-tier indexes
+
+These indexes on the per-webhook event databases are declared in the model
+tags, so `AutoMigrate` creates them on a fresh and on an existing database:
+
+| Table              | Columns                     | Serves |
+| ------------------ | --------------------------- | ------ |
+| `deliveries`       | `status`, `deleted_at`      | Startup recovery, the retry and pending sweeps every 60 seconds and the queue-depth sampler every 30 seconds, which select deliveries by status |
+| `deliveries`       | `event_id`, `deleted_at`    | The event log, which loads each event's deliveries, and retention, which selects and deletes the deliveries of expired events |
+| `delivery_results` | `delivery_id`, `deleted_at` | The event log, which loads the attempts of a page's deliveries, and retention, which deletes the attempts of expired events |
+| `events`           | `deleted_at`, `created_at`  | Retention, which selects expired events by age |
+| `events`           | `created_at`                | Retention's delete of the expired events themselves |
+
+GORM's soft delete adds `deleted_at IS NULL` to these queries; retention's
+deletes leave it out, but their lookups of expired rows keep it. SQLite keeps
+no statistics on these tables, and without them it rates the `deleted_at`
+index, which every live row matches, above an index on a column matched
+against several values or compared with `<`. So every index but the last also
+covers `deleted_at`. It comes second, so that retention's deletes can use the
+index without it, except in `events`, where `created_at` is compared with `<`
+and SQLite narrows by a `<` only on the last column it uses.
+
 #### Common Fields

 Every entity except `Setting` includes these fields from `BaseModel`.
@@ -1308,6 +1775,11 @@ webhooker uses **separate SQLite database files**: a main application
 database for configuration data and per-webhook databases for event
 storage. All database files live in the `DATA_DIR` directory.

+Every one of them is created `0600`, and so is each `-wal` and `-shm`
+sidecar. See
+[Running with Docker](#running-with-docker) for what that does and
+does not protect.
+
 **Main Application Database** (`{DATA_DIR}/webhooker.db`) — stores
 configuration and application state:

@@ -1349,10 +1821,12 @@ This separation provides:
  only, or disables cleanup entirely when set to `0` (retain forever).
 - **Performance** — each webhook's database has its own page cache and
  its own lock, so concurrent event ingestion across webhooks won't
-  contend. No write-ahead log is involved: both DSNs are
-  `file:{path}?cache=shared&mode=rwc` and no `journal_mode` pragma is
-  ever issued, so every database runs on SQLite's default rollback
-  journal.
+  contend. Every database — main, per-webhook, and archive — is opened
+  through one code path (`internal/database/sqlite_open.go`) in WAL
+  journal mode, with a 10-second busy timeout, `BEGIN IMMEDIATE`
+  transactions, and a bounded connection pool. Under WAL a reader never
+  blocks a writer, so an operator reading a database does not stall
+  event ingestion into it.

 The **database target type** builds on this architecture to provide
 long-term archiving, separate from the per-webhook event database (which
@@ -1543,7 +2017,7 @@ rescans the database anyway).
 | ----------- | -------- |
 | **Closed**  | Normal operation. Deliveries flow through. Consecutive failures are counted. |
 | **Open**    | Target appears down. Deliveries are skipped and rescheduled for after the cooldown. |
-| **Half-Open** | Cooldown expired. One probe delivery is allowed to test if the target has recovered. |
+| **Half-Open** | Cooldown expired. One probe delivery is allowed to test if the target has recovered. Other deliveries are rescheduled for one whole cooldown later. |

 **Transitions:**

@@ -1581,7 +2055,9 @@ operations), and log targets (stdout) do not use circuit breakers.
 When a circuit is open and a new delivery arrives, the engine marks the
 delivery as `retrying` and schedules a retry timer for after the
 remaining cooldown period. This ensures no deliveries are lost — they're
-just delayed until the target is healthy again.
+just delayed until the target is healthy again. A delivery already in
+`retrying` keeps that status without another database write each time
+the breaker turns it away.

 ### Metrics

@@ -1595,7 +2071,7 @@ arriving and being stored, they are just not getting anywhere.
 | Metric | Type | Meaning |
 | ------ | ---- | ------- |
 | `webhooker_events_received_total` | counter | Events received and durably stored. Compare against the delivery counters on one dashboard |
-| `webhooker_delivery_attempts_total` | counter | Delivery attempts actually dispatched to a target. A delivery an open circuit breaker refused is not one: it is counted as a retry instead |
+| `webhooker_delivery_attempts_total` | counter | Delivery attempts actually dispatched to a target. A delivery a circuit breaker refused is not one: it is counted as a retry instead, but only when the refusal moves it into `retrying` |
 | `webhooker_deliveries_succeeded_total` | counter | Deliveries that reached `delivered` |
 | `webhooker_deliveries_failed_total` | counter | Deliveries that failed terminally and will not be retried |
 | `webhooker_delivery_retries_total` | counter | Deliveries put back into `retrying` |
@@ -2212,7 +2688,7 @@ abuse limit later; they are tracked as future work.
 | ------ | --------------------------- | ----------- |
 | `GET`  | `/`                         | Root redirect, 303 (authenticated → `/sources`, unauthenticated → `/pages/login`) |
 | `GET`  | `/.well-known/healthcheck`  | Health check (JSON: `status`, `now`, `uptimeSeconds`, `uptimeHuman`, `version`, `appname`, `maintenanceMode`) |
-| any    | `/s/*`                      | Static file serving (embedded CSS, JS). Mounted for every method, not just `GET`/`HEAD`: chi's `Mount` registers all methods and `http.FileServer` special-cases only `HEAD` (by omitting the body), so a `POST` or `DELETE` to an asset is answered `200` with the file. Pinned by `TestStaticServesEveryMethod` |
+| `GET`, `HEAD` | `/s/*`              | Static file serving (embedded CSS, JS). `GET` and `HEAD` only — `POST`, `PUT`, `PATCH`, `DELETE`, `OPTIONS`, `TRACE` and `CONNECT` are answered `405 Method Not Allowed` with `Allow: GET, HEAD`. Any other method (such as `PROPFIND`) is refused by chi before it reaches this route, and gets `405` without an `Allow` header. Pinned by `TestStaticServesOnlyGetAndHead` |
 | `POST` | `/webhook/{uuid}`           | Webhook receiver endpoint. `POST` only — every other method is answered `405 Method Not Allowed` with `Allow: POST`. Rate limited (see [Rate Limiting](#rate-limiting)) |

 #### Authentication Endpoints
@@ -2237,12 +2713,15 @@ abuse limit later; they are tracked as future work.
 | `POST` | `/source/{id}/edit`      | Edit webhook submission |
 | `POST` | `/source/{id}/delete`    | Delete webhook |
 | `GET`  | `/source/{id}/logs`      | Webhook event logs |
+| `GET`  | `/source/{id}/logs/{eventID}/body` | Download an event's full stored body. The log page renders each body only up to its cap, so this is the only route that serves a whole one; it is offered wherever a body is shown truncated |
 | `POST` | `/source/{id}/deliveries/{deliveryID}/replay` | Replay a finished delivery: creates a new delivery for the same event against the target's current configuration (30 per minute per bucket, then `429`) |
 | `POST` | `/source/{id}/events/{eventID}/resubmit` | Resubmit a stored event: creates a new event copying it and fans that out to every currently active target (30 per minute per bucket, then `429`) |
 | `POST` | `/source/{id}/entrypoints` | Add entrypoint to webhook |
 | `POST` | `/source/{id}/entrypoints/{entrypointID}/delete` | Delete an entrypoint |
 | `POST` | `/source/{id}/entrypoints/{entrypointID}/toggle` | Enable or disable an entrypoint |
 | `POST` | `/source/{id}/targets`   | Add target to webhook |
+| `GET`  | `/source/{id}/targets/{targetID}/edit` | Edit target form. The one page that renders a target's destination URL and header values in full, rather than masked |
+| `POST` | `/source/{id}/targets/{targetID}/edit` | Edit target submission |
 | `POST` | `/source/{id}/targets/{targetID}/delete` | Delete a target |
 | `POST` | `/source/{id}/targets/{targetID}/toggle` | Enable or disable a target |

@@ -2281,6 +2760,8 @@ webhooker/
 ├── internal/
 │   ├── banner/
 │   │   └── banner.go          # Ruled block for the one credential shown in the clear
+│   ├── ciscript/
+│   │   └── doc.go             # Tests for the CI shell scripts in script/; no runtime code
 │   ├── resetpw/
 │   │   └── resetpw.go         # `webhooker resetpw`: set an account's password, stopped deployments only
 │   ├── config/
@@ -2350,13 +2831,17 @@ webhooker/
 │   │   ├── ratelimit.go       # Per-IP rate limiting middleware (go-chi/httprate)
 │   │   ├── loginguard.go      # Login failure counters and the Argon2id verification semaphore
 │   │   └── testing.go         # NewForTest: Middleware without the fx lifecycle
+│   ├── reqtls/
+│   │   └── reqtls.go          # IsTLS: the one TLS predicate, r.TLS or X-Forwarded-Proto
 │   ├── server/
 │   │   ├── server.go          # Server struct, fx lifecycle, signal handling
 │   │   ├── http.go            # HTTP server setup with timeouts
 │   │   └── routes.go          # All route definitions
-│   └── session/
-│       ├── session.go         # Cookie-based session management
-│       └── testing.go         # NewForTest: Session without the fx lifecycle
+│   ├── session/
+│   │   ├── session.go         # Cookie-based session management
+│   │   └── testing.go         # NewForTest: Session without the fx lifecycle
+│   └── versionscript/
+│       └── doc.go             # Tests for script/version and the build files that use it
 ├── static/
 │   ├── static.go              # //go:embed directive
 │   ├── css/input.css          # Tailwind input, source for tailwind.css (make css)
@@ -2469,6 +2954,10 @@ check, see [The login endpoint](#the-login-endpoint).

 ### Authentication

+- **Webhook receiver:** the entrypoint UUID in the URL, and nothing
+  else. No shared secret, no HMAC signature, no token header, and none
+  will be added — see
+  [The entrypoint URL is the authentication secret](#the-entrypoint-url-is-the-authentication-secret).
 - **Web UI:** Cookie-based sessions using gorilla/sessions with
  encrypted cookies. Sessions are configured with HttpOnly, SameSite
  Lax, and Secure whenever the request is on TLS — the flag follows the
@@ -2508,7 +2997,8 @@ check, see [The login endpoint](#the-login-endpoint).
  mode
 - **The entrypoint URL is the receiver's only credential.** Nothing
  about an inbound request is verified; possession of the UUID
-  authorises submission (see
+  authorises submission, and no shared secret or signature check will
+  be added alongside it (see
  [The entrypoint URL is the authentication secret](#the-entrypoint-url-is-the-authentication-secret))
 - **SSRF prevention** for HTTP delivery targets: private/reserved IP
  ranges (RFC 1918, loopback, link-local, cloud metadata) are blocked
@@ -2548,7 +3038,11 @@ check, see [The login endpoint](#the-login-endpoint).
 - Prometheus metrics behind basic auth
 - Static assets embedded in binary (no filesystem access needed at
  runtime)
- Container runs as non-root user (UID 1000)
+- The app runs as the non-root `webhooker` user (UID 1000) in the
+  container. The image sets no `USER`, so these run as root: the
+  `ENTRYPOINT` script, which sets the data directory's owner and mode
+  before the app starts; the image's health check; and `docker exec`,
+  unless given `--user`
 - GORM soft deletes on every entity that carries `BaseModel`, which is
  all of them but `Setting` (data preserved for audit)

@@ -2564,7 +3058,8 @@ each hook. The order, read off the fx stop-hook log:
 3. `server` — the HTTP drain, bounded separately by
   `server.ShutdownTimeout` (**3 seconds**), then a Sentry flush if
   `SENTRY_DSN` is set
-4. `delivery.Engine`
+4. `delivery.Engine` — waits for its workers, then closes the archive
+   databases
 5. `healthcheck`
 6. `WebhookDBManager`
 7. the database close
@@ -2674,10 +3169,13 @@ version is fixed independently of the compiler's:
   `GO_LDFLAGS`, so neither can drop the `-X` that stamps the version.
   The version arrives as the `VERSION` build arg, since the context
   has no `.git` (see [Version stamping](#version-stamping)).
-3. **Runtime stage** (`alpine:3.21`) — copies the static binary,
-   creates the `/var/lib/webhooker` directory for all SQLite databases,
-   runs as the non-root `webhooker` user (UID 1000), exposes port 8080,
-   and includes a health check against `/.well-known/healthcheck`.
+3. **Runtime stage** (`alpine:3.21`) — copies the static binary and
+   `deploy/docker-entrypoint.sh`, creates the `/var/lib/webhooker`
+   directory for all SQLite databases, exposes port 8080, and includes
+   a health check against `/.well-known/healthcheck`. It sets no
+   `USER`: the `ENTRYPOINT` script starts as root, sets the data
+   directory's owner and mode, and runs the app as the non-root
+   `webhooker` user (UID 1000) through `su-exec`.

 The lint stage invokes `golangci-lint` directly rather than `make lint`:
 it is already the pinned linter image, and `make lint` builds
@@ -2756,3 +3254,5 @@ MIT
 ## Author

 [@sneak](https://sneak.berlin)
+
+
@@ -18,29 +18,27 @@ Issue branches do NOT touch this file — the manager maintains it on

 # Status

-1.0.0 is complete: 55 closed, 0 open. `next` (6874059) is 62 commits
-ahead of `main` and a strict fast-forward. No git tags exist yet.
+The milestone (https://git.eeqj.de/sneak/webhooker/milestone/9) is the
+authoritative list, and the only place to read a count or a state of
+play from. This file records where the project is, not what is in
+flight: a sentence whose truth depends on a branch being unmerged is
+wrong the moment it merges, and this file has been wrong that way
+before.

-The bar was not "the milestone is empty" but "sneak can deploy this and
-use it in low-volume production". Every gap the deployability audit
-named against that bar is now closed:
+The durability defect that held the tag has landed
+(https://git.eeqj.de/sneak/webhooker/issues/256, commit `8d64259`).
+Every SQLite handle opens with WAL journaling and a busy timeout, a
+bookkeeping write that fails leaves its delivery in a recoverable
+state rather than a lying one, and recovery skips a delivery that
+already has a successful result row. Final pre-tag verification
+exercised it and confirmed it holds. Whatever the milestone still
+shows open is what remains before `v1.0.0`.

- `DATA_DIR` locking, so two instances cannot both deliver
-  (https://git.eeqj.de/sneak/webhooker/issues/201)
- shutdown on listener failure, rather than a live non-serving process
-  (https://git.eeqj.de/sneak/webhooker/issues/200)
- inbound signature verification
-  (https://git.eeqj.de/sneak/webhooker/issues/67)
- per-attempt delivery detail in the event log
-  (https://git.eeqj.de/sneak/webhooker/issues/202)
- replay of a terminally failed delivery
-  (https://git.eeqj.de/sneak/webhooker/issues/203)
- `ALLOWED_EGRESS_CIDRS`, an allowlist escape hatch for the SSRF guard
-  (https://git.eeqj.de/sneak/webhooker/issues/204)
- the three credential exposures
-  (https://git.eeqj.de/sneak/webhooker/issues/205,
-  https://git.eeqj.de/sneak/webhooker/issues/206,
-  https://git.eeqj.de/sneak/webhooker/issues/207)
+Delivery is at-least-once by design, not by accident: a send whose
+result row does not land is attempted again, so a receiver can see a
+duplicate. That is deliberate — the alternative is a silent lost
+delivery — and the README says so under Rationale. It is not a defect
+to re-file.

 One caveat on reading a green check: a docs-only commit deliberately
 replays from the layer cache
@@ -50,23 +48,129 @@ commit invalidates the `COPY` layer and genuinely executes.

 # Next Step

-Merge the milestone PR (https://git.eeqj.de/sneak/webhooker/pulls/111)
-and tag `v1.0.0`. It is `merge-ready` and assigned to sneak; nothing
-else gates it.
-
-Post-1.0 follow-ups are open, none blocking the tag:
-https://git.eeqj.de/sneak/webhooker/issues/245,
-https://git.eeqj.de/sneak/webhooker/issues/246,
-https://git.eeqj.de/sneak/webhooker/issues/247 and
-https://git.eeqj.de/sneak/webhooker/issues/248. Also still open and
-unmilestoned: https://git.eeqj.de/sneak/webhooker/issues/193 (a design
-question, not a defect), https://git.eeqj.de/sneak/webhooker/issues/198
-(`make test` is past the org 20s target) and
-https://git.eeqj.de/sneak/webhooker/issues/212 (encrypting target config
-at rest).
+Clear the rest of the open 1.0.0 milestone
+(https://git.eeqj.de/sneak/webhooker/milestone/9) and tag `v1.0.0`.
+Merging `next` into `main` is a separate act from tagging and waits on
+neither of those: `next` is kept mergeable at all times, which is the
+point of the branch.

 # Completed Steps

+- 2026-08-24 Bind the plaintext HTTP listener deliberately, via
+  `BIND_ADDRESS` defaulting to `127.0.0.1`, and document the
+  reverse-proxy deployment. A hostname, an empty value or a value
+  carrying a port is a startup error, and the `Dockerfile` sets
+  `0.0.0.0` because a loopback bind inside a container is unreachable
+  (https://git.eeqj.de/sneak/webhooker/issues/268). The same commit
+  removed the shutdown race: `httpServer` is built in the constructor
+  rather than assigned from the serving goroutine, which orders the
+  write before every fx hook and rules out the nil dereference a
+  SIGTERM arriving first would have caused, and `sentryEnabled` is an
+  `atomic.Bool` (https://git.eeqj.de/sneak/webhooker/issues/226)
+- 2026-08-24 Remove inbound request signature verification. The
+  entrypoint UUID is the authentication secret, so the per-entrypoint
+  shared secret, the `internal/signature` package, the receiver check,
+  the model fields and the forms are all gone. This reverses the
+  feature that landed earlier in the same milestone
+  (https://git.eeqj.de/sneak/webhooker/issues/67,
+  https://git.eeqj.de/sneak/webhooker/issues/279)
+- 2026-08-24 Stamp the build version into the binary and render it in
+  the UI footer. `script/version` is the single source — `$VERSION`,
+  else `git describe --tags --always --dirty`, else `unknown` — so a
+  `make build` binary and a `make docker` image from one checkout
+  report the same thing, and nothing in it varies between two builds
+  of the same commit, which the release gate's byte-identical
+  assertion would catch
+  (https://git.eeqj.de/sneak/webhooker/issues/253)
+- 2026-08-24 Derive cookie `Secure` and CSRF strictness from the
+  request transport rather than from `WEBHOOKER_ENVIRONMENT`. Behind a
+  real TLS proxy with the environment left at its `dev` default, the
+  session cookie silently lost `Secure` while the CSRF cookie on the
+  same response kept it. `X-Forwarded-Proto` is now matched
+  case-insensitively on its first comma-separated element, so `HTTPS`
+  and `https, http` no longer fall to the relaxed CSRF path
+  (https://git.eeqj.de/sneak/webhooker/issues/269)
+- 2026-08-24 Roll back a failed webhook deletion instead of committing
+  it. A failing delete committed whatever had already succeeded,
+  hard-deleted the per-webhook event database anyway, and redirected as
+  though it had worked — orphaned config plus permanently destroyed
+  history, reported as success. All three delete positions now roll
+  back with the event database intact
+  (https://git.eeqj.de/sneak/webhooker/issues/262)
+- 2026-08-24 Name a deleted target on its historical deliveries, marked
+  `(deleted)`, rather than leaving the event log unable to say where a
+  delivery went. A deleted target's credentials stay masked exactly as
+  a live one's, and it cannot become deliverable again through the
+  receiver, resubmit, replay, the edit form or the toggle
+  (https://git.eeqj.de/sneak/webhooker/issues/211)
+- 2026-08-24 Bound both request-controlled `/metrics` label dimensions,
+  so the unauthenticated receiver is no longer a memory-exhaustion
+  vector: `handler` carries the chi route pattern, and `method` folds
+  anything chi cannot route onto a single `(unmatched)` sentinel. Both
+  were reproduced before the fix — 300 random method tokens took the
+  series count from 106 to 7,631, and path flooding reached 62,532 —
+  and a label audit across a live scrape found no third unbounded
+  dimension (https://git.eeqj.de/sneak/webhooker/issues/254,
+  https://git.eeqj.de/sneak/webhooker/issues/261)
+- 2026-08-24 Validate `max_retries` on both target forms. `abc`, `2.7`
+  and `-5` silently became 0 — fire-and-forget — including on the edit
+  path, where it destroyed a working value, and `999999999` stored
+  verbatim. The ceiling of 20 is the `max` both templates already
+  declared (https://git.eeqj.de/sneak/webhooker/issues/221)
+- 2026-08-24 Resubmit a stored event as a new undelivered event, so a
+  backend under development can be tested against real captured
+  traffic. Per-delivery replay cannot serve that: it re-sends one
+  finished delivery to its own original target, and a target created
+  for a dev backend has no prior delivery to replay. Resubmit
+  re-injects the stored event at the top of the receiver path and fans
+  it out to whatever targets are active now
+  (https://git.eeqj.de/sneak/webhooker/issues/250)
+- 2026-08-20 Take an exclusive lock on `DATA_DIR` at startup, so two
+  instances on one directory cannot both deliver
+  (https://git.eeqj.de/sneak/webhooker/issues/201)
+- 2026-08-20 Shut down the app when the HTTP listener fails. The
+  `OnStart` hook returned as soon as the serving goroutine was
+  spawned, so a failed listen left fx reporting RUNNING and a live
+  process with nothing bound — invisible to systemd and Docker restart
+  policies (https://git.eeqj.de/sneak/webhooker/issues/200)
+- 2026-08-20 Stop target credentials leaking into the per-webhook event
+  databases (https://git.eeqj.de/sneak/webhooker/issues/206), log SQL
+  with placeholders rather than bound values
+  (https://git.eeqj.de/sneak/webhooker/issues/207), and fail loudly on
+  half-set metrics auth credentials
+  (https://git.eeqj.de/sneak/webhooker/issues/205)
+- 2026-08-20 Read queue depths with `Find`, not `Scan`. `Scan` swaps
+  GORM's own trace recorder in for the logging adapter, and that
+  recorder does not implement `gorm.ParamsFilter`, so those statements
+  logged their bound values interpolated and bypassed the suppression
+  above. The two units gated green against a `next` that lacked the
+  other, and `next` went red when both landed
+  (https://git.eeqj.de/sneak/webhooker/issues/234)
+- 2026-08-20 Render per-attempt delivery detail in the event log
+  (https://git.eeqj.de/sneak/webhooker/issues/202) and add replay of a
+  terminally failed delivery
+  (https://git.eeqj.de/sneak/webhooker/issues/203)
+- 2026-08-20 Expose delivery metrics on `/metrics`
+  (https://git.eeqj.de/sneak/webhooker/issues/209) and document the
+  backup, restore and upgrade procedures
+  (https://git.eeqj.de/sneak/webhooker/issues/210)
+- 2026-08-20 Add a `webhooker resetpw` subcommand and a bootstrap
+  banner. The admin bootstrap password was printed once among roughly
+  45 fx lines, and under `docker run -d` went to container logs subject
+  to rotation; there was no reset path at all, so recovery meant
+  hand-deleting the users row, documented nowhere. The password is read
+  from stdin or generated, never from argv where `/proc` would publish
+  it (https://git.eeqj.de/sneak/webhooker/issues/208)
+- 2026-08-20 Add `ALLOWED_EGRESS_CIDRS`, an allowlist-only escape hatch
+  for the SSRF guard, so a self-hosted proxy can forward into the
+  operator's own network. The guard's always-blocked set cannot be
+  reopened by configuration
+  (https://git.eeqj.de/sneak/webhooker/issues/204)
+- 2026-08-20 Harden operator-set target headers, which were carried
+  unsafely across a redirect
+  (https://git.eeqj.de/sneak/webhooker/issues/233)
+- 2026-08-20 Add a target edit form with headers and timeout fields
+  (https://git.eeqj.de/sneak/webhooker/issues/127)
 - 2026-08-18 Raise `script/test`'s per-package timeout from 30s to 90s,
  matching the org-wide backstop. `go test` applies `-timeout` per
  package, and `internal/handlers` had grown past the old budget: a
@@ -299,9 +403,6 @@ at rest).
 - Per-webhook rate limiting in the receiver handler (per-webhook config
  plus handler enforcement; global limits must not apply to receiver
  endpoints)
- Stripe HMAC signature verification. The GitHub and GitLab schemes
-  landed with inbound verification
-  (https://git.eeqj.de/sneak/webhooker/issues/67)
 - API key authentication for programmatic access (APIKey model exists;
  Bearer token middleware does not)
 - REST API v1
@@ -0,0 +1,107 @@
+package main
+
+import (
+	"bytes"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/webhooker/internal/config"
+)
+
+// dotEnvKey is a throwaway variable name these tests write and read,
+// so they cannot disturb real configuration.
+const dotEnvKey = "WEBHOOKER_TEST_DISPATCH_VALUE"
+
+// writeDotEnvInWorkingDir puts contents in a .env file in a fresh
+// temporary directory and moves the process there.
+//
+// The callers are deliberately not parallel and must stay that way:
+// t.Chdir moves the whole process. Go releases parallel tests only
+// after every sequential test in the package has finished, so nothing
+// else runs while these do.
+func writeDotEnvInWorkingDir(t *testing.T, contents string) {
+	t.Helper()
+
+	dir := t.TempDir()
+	require.NoError(t, os.WriteFile(
+		filepath.Join(dir, config.DotEnvPath),
+		[]byte(contents), 0o600,
+	))
+	t.Chdir(dir)
+}
+
+// TestDispatch_MalformedDotEnvRefuses pins the second half of the
+// defect. godotenv applies nothing at all when a file will not parse,
+// so one mistyped line used to revert every variable in it to its
+// default and start the server anyway, with no log line naming the
+// file. The refusal has to arrive before any subcommand runs, which
+// is why `help` — the one subcommand that touches nothing — is still
+// refused here.
+//
+//nolint:paralleltest // t.Chdir moves the whole process.
+func TestDispatch_MalformedDotEnvRefuses(t *testing.T) {
+	writeDotEnvInWorkingDir(t, "PORT 19615\n")
+
+	var stdout, stderr bytes.Buffer
+
+	code := dispatch(
+		[]string{helpCommand}, strings.NewReader(""), &stdout, &stderr,
+	)
+
+	require.Equal(t, 1, code, "a broken .env must exit non-zero")
+	assert.Contains(
+		t, stderr.String(), config.DotEnvPath,
+		"the refusal must name the file",
+	)
+	assert.Empty(
+		t, stdout.String(),
+		"the subcommand must not have run",
+	)
+}
+
+// TestDispatch_LoadsDotEnvBeforeSubcommands pins the ordering the
+// godotenv/autoload import used to provide for free. It ran in an
+// init(), so .env was in the environment before anything read it —
+// including config.DataDir, which both the DATA_DIR lock and resetpw
+// call outside the fx graph. Loading any later would let a .env that
+// sets DATA_DIR lock one directory while the config opened databases
+// in another.
+func TestDispatch_LoadsDotEnvBeforeSubcommands(t *testing.T) {
+	t.Setenv(dotEnvKey, "placeholder")
+	require.NoError(t, os.Unsetenv(dotEnvKey))
+
+	writeDotEnvInWorkingDir(t, dotEnvKey+"=from-dot-env\n")
+
+	var stdout, stderr bytes.Buffer
+
+	code := dispatch(
+		[]string{helpCommand}, strings.NewReader(""), &stdout, &stderr,
+	)
+
+	require.Equal(t, 0, code)
+	assert.Equal(
+		t, "from-dot-env", os.Getenv(dotEnvKey),
+		"the file must be applied before the subcommand runs",
+	)
+}
+
+// TestDispatch_MissingDotEnvIsFine pins the case most deployments are
+// in: no .env at all, which must stay a normal start.
+//
+//nolint:paralleltest // t.Chdir moves the whole process.
+func TestDispatch_MissingDotEnvIsFine(t *testing.T) {
+	t.Chdir(t.TempDir())
+
+	var stdout, stderr bytes.Buffer
+
+	code := dispatch(
+		[]string{helpCommand}, strings.NewReader(""), &stdout, &stderr,
+	)
+
+	require.Equal(t, 0, code)
+	assert.Empty(t, stderr.String())
+}
@@ -54,6 +54,11 @@ const stopTimeout = 5 * time.Second
 // caller can tell "called wrong" from "declined".
 const exitUsage = 2

+// helpCommand is the subcommand that prints usage. The flag spellings
+// beside it in the switch are aliases; this is the name the usage text
+// documents and the one tests invoke.
+const helpCommand = "help"
+
 // Build-time variables set via -ldflags.
 //
 //nolint:gochecknoglobals // Build-time variables injected by the linker.
@@ -75,11 +80,27 @@ func main() {
 // every existing deployment invoke; that path is unchanged, including
 // where the DATA_DIR lock is taken relative to building the fx graph
 // and how fx propagates a non-zero exit itself.
+//
+// The optional .env file is read here, before any subcommand and so
+// before anything reads the environment — config.DataDir, which both
+// the DATA_DIR lock and resetpw call outside the fx graph, above all.
+// It used to be read from an init() in internal/config, which put it
+// earlier still but threw the error away: a single malformed line
+// applied none of the file and said nothing about it. A file that is
+// not there stays fine, since .env is optional and most deployments
+// do not have one.
 func dispatch(
 	args []string,
 	stdin io.Reader,
 	stdout, stderr io.Writer,
 ) int {
+	err := config.LoadDotEnv()
+	if err != nil {
+		_, _ = fmt.Fprintf(stderr, "%s: %v\n", appname, err)
+
+		return 1
+	}
+
 	if len(args) == 0 {
 		return run(stderr)
 	}
@@ -87,7 +108,7 @@ func dispatch(
 	switch args[0] {
 	case resetpw.Name:
 		return resetpw.Run(args[1:], stdin, stdout, stderr)
-	case "help", "-h", "-help", "--help":
+	case helpCommand, "-h", "-help", "--help":
 		usage(stdout)

 		return 0
@@ -121,7 +121,7 @@ func TestDispatch_Help(t *testing.T) {
 	var stdout, stderr bytes.Buffer

 	code := dispatch(
-		[]string{"help"}, strings.NewReader(""), &stdout, &stderr,
+		[]string{helpCommand}, strings.NewReader(""), &stdout, &stderr,
 	)

 	require.Equal(t, 0, code)
@@ -0,0 +1,22 @@
+#!/bin/sh
+# deploy/docker-entrypoint.sh: the image's ENTRYPOINT. A bind-mounted
+# data directory keeps its owner from the host, often root, and the app
+# could not write to it. Started as root, this creates DATA_DIR if
+# needed, gives it and everything in it to webhooker, sets its mode, and
+# runs the command as webhooker, so the app never runs as root. Started
+# as another user, it only runs the command.
+set -eu
+
+main() {
+    if [ "$(id -u)" != 0 ]; then
+        exec "$@"
+    fi
+
+    dir="${DATA_DIR:-/var/lib/webhooker}"
+    mkdir -p "$dir"
+    find "$dir" ! -user webhooker -exec chown -h webhooker:webhooker {} +
+    chmod 750 "$dir"
+    exec su-exec webhooker "$@"
+}
+
+main "$@"
@@ -4,6 +4,7 @@ package config
 import (
 	"errors"
 	"fmt"
+	"io/fs"
 	"log/slog"
 	"net/netip"
 	"os"
@@ -11,13 +12,11 @@ import (
 	"strings"
 	"time"

+	"github.com/getsentry/sentry-go"
+	"github.com/joho/godotenv"
 	"go.uber.org/fx"
 	"sneak.berlin/go/webhooker/internal/globals"
 	"sneak.berlin/go/webhooker/internal/logger"
-
-	// Populates the environment from a ./.env file automatically for
-	// development configuration. Kept in one place only (here).
-	_ "github.com/joho/godotenv/autoload"
 )

 const (
@@ -33,6 +32,34 @@ const (
 	// defaultPort is the default HTTP listen port.
 	defaultPort = 8080

+	// defaultBindAddress is the interface the plaintext HTTP
+	// listener claims when BIND_ADDRESS is unset.
+	//
+	// Loopback, because the listener speaks cleartext and serves
+	// both the admin UI and the unauthenticated receiver: a
+	// wildcard default publishes them on every interface of every
+	// host that never configured anything, which is the failure
+	// this default exists to prevent. Reaching webhooker from off
+	// the host is then a deliberate act — a reverse proxy in front
+	// of it, or an explicit BIND_ADDRESS.
+	//
+	// This is the binary's default only. The Dockerfile ships
+	// ENV BIND_ADDRESS=0.0.0.0, so a container deployment needs
+	// nothing set and is unaffected by this constant. The two
+	// differ because they answer different questions: a container's
+	// network namespace is already the boundary this default is
+	// reaching for, so binding every address inside it exposes
+	// nothing, and what decides exposure there is the publish flag
+	// (-p 127.0.0.1:8080:8080). A loopback bind inside a container
+	// buys no security and makes the process unreachable through
+	// its own published port.
+	//
+	// The split is expressed as two explicit defaults rather than
+	// container auto-detection, because a heuristic that guesses
+	// wrong opens the cleartext port exactly where nobody is
+	// looking.
+	defaultBindAddress = "127.0.0.1"
+
 	// defaultRetentionSweepInterval is how often the retention
 	// reaper deletes events older than each webhook's RetentionDays.
 	defaultRetentionSweepInterval = time.Hour
@@ -56,6 +83,12 @@ const (
 	// IPv6 prefix spends on the ::ffff:0:0/96 wrapper, so a /104
 	// covers the same addresses as an IPv4 /8.
 	mappedV4Offset = 96
+
+	// DotEnvPath is the optional file of KEY=value lines read into the
+	// environment at startup, relative to the process working
+	// directory. Exported so that documentation and tests name the
+	// same path the loader opens.
+	DotEnvPath = ".env"
 )

 // ErrInvalidEnvironment is returned when WEBHOOKER_ENVIRONMENT
@@ -75,6 +108,19 @@ var ErrInvalidPort = errors.New("invalid port")
 // nor a bare IP address.
 var ErrInvalidCIDR = errors.New("invalid CIDR")

+// ErrInvalidBindAddress is returned when BIND_ADDRESS is set to
+// something that is not an IP address literal.
+var ErrInvalidBindAddress = errors.New("invalid bind address")
+
+// ErrInvalidSentryDSN is returned when SENTRY_DSN is set to something
+// the Sentry SDK cannot parse as a DSN.
+var ErrInvalidSentryDSN = errors.New("invalid Sentry DSN")
+
+// ErrDotEnvUnreadable is returned when the optional .env file exists
+// but cannot be read or parsed. A file that is not there is not an
+// error; a file that is there and broken is.
+var ErrDotEnvUnreadable = errors.New("unreadable .env file")
+
 // ErrIncompleteMetricsAuth is returned when exactly one of
 // METRICS_USERNAME and METRICS_PASSWORD carries a value. Neither
 // fallback is acceptable: serving /metrics on the username alone
@@ -105,6 +151,13 @@ type Config struct {
 	Port            int
 	SentryDSN       string

+	// BindAddress is the IP address the plaintext HTTP listener
+	// binds, as an address literal. It defaults to
+	// defaultBindAddress and is never empty: an empty string would
+	// mean the wildcard to net.Listen, which is the opposite of the
+	// default this ships.
+	BindAddress string
+
 	// RetentionSweepInterval is how often the retention reaper runs.
 	// Always positive: it becomes a time.NewTicker period.
 	RetentionSweepInterval time.Duration
@@ -139,9 +192,10 @@ type Config struct {
 	// alwaysBlockedNetworks stays blocked no matter what is listed
 	// here. That set is link-local plus the cloud metadata
 	// endpoints outside it that disclose credentials or user data
-	// at a provider-fixed address; it is not exhaustive of every
-	// cloud's metadata address. See alwaysBlockedNetworks for the
-	// authoritative list and the criterion it is built from.
+	// at a provider-fixed, non-public address; it is not
+	// exhaustive of every cloud's metadata address. See
+	// alwaysBlockedNetworks for the authoritative list and the
+	// criterion it is built from.
 	AllowedEgressCIDRs []netip.Prefix

 	params *ConfigParams
@@ -173,12 +227,62 @@ func (c *Config) MetricsAuthEnabled() bool {
 	return c.MetricsUsername != "" && c.MetricsPassword != ""
 }

+// SentryEnabled reports whether error reporting is shipped to Sentry.
+// It is the only answer to that question in the codebase: the SDK
+// initialisation, the sentryhttp middleware registration and the
+// startup log's sentryEnabled field all read this one method, so the
+// log cannot report reporting as on while nothing is sending.
+//
+// A non-empty DSN is enough because loadFromEnv already parsed it with
+// the SDK's own parser and refused to build a Config around one the
+// SDK would reject, and because initialising the SDK with a DSN that
+// parsed and failed anyway aborts the process rather than leaving this
+// true and the client absent.
+func (c *Config) SentryEnabled() bool {
+	return c.SentryDSN != ""
+}
+
 // envString returns the value of the named environment variable,
 // or an empty string if not set.
 func envString(key string) string {
 	return os.Getenv(key)
 }

+// LoadDotEnv reads DotEnvPath into the environment when that file is
+// present, and reports a file that is present but broken.
+//
+// It has to run before anything reads the environment, so that every
+// reader agrees on what the environment holds — the DATA_DIR lock
+// taken before the fx graph exists as much as loadFromEnv itself. A
+// variable already set in the real environment wins: godotenv never
+// overwrites one.
+//
+// A missing file is not an error. It is a development convenience and
+// most deployments set the environment directly.
+//
+// Any other failure is. godotenv parses the whole file before setting
+// anything, so a single malformed line applies none of it: every
+// variable in the file silently reverts to its default, which defeats
+// the fail-loud guarantee for all of them at once.
+func LoadDotEnv() error {
+	return loadDotEnvFile(DotEnvPath)
+}
+
+// loadDotEnvFile is LoadDotEnv over a named file, so tests can point
+// at a temporary one instead of the process working directory.
+func loadDotEnvFile(path string) error {
+	err := godotenv.Load(path)
+	if err == nil || errors.Is(err, fs.ErrNotExist) {
+		return nil
+	}
+
+	return fmt.Errorf(
+		"%w: %s: %w; nothing in it was applied, so fix the file or "+
+			"remove it",
+		ErrDotEnvUnreadable, path, err,
+	)
+}
+
 // DataDir resolves DATA_DIR, applying DefaultDataDir when it is unset
 // or empty. It is exported so that entry points which must act on the
 // data directory before the fx graph exists — taking the exclusive
@@ -387,6 +491,77 @@ func envPrefixList(key string) ([]netip.Prefix, error) {
 	return prefixes, nil
 }

+// envBindAddress returns the value of the named environment variable
+// parsed as an IP address literal. An unset (or empty, or
+// whitespace-only) value yields defaultValue.
+//
+// Only literals are accepted: no hostname is resolved, so `localhost`
+// is an error rather than a DNS lookup at startup whose answer could
+// be either loopback family, could change under the process, and
+// could return several addresses of which only one would be bound. A
+// value with a port in it (`127.0.0.1:8080`) is likewise an error —
+// the port is PORT's business, and silently accepting it would bind
+// something other than what was asked for.
+//
+// A set value that is not a literal is a hard error naming the key
+// and the bad value, so startup fails loudly rather than falling back
+// to a default the operator plainly did not want. A literal that is
+// not an address of this host parses here and fails at listen time
+// instead, which ends the process non-zero.
+func envBindAddress(key, defaultValue string) (string, error) {
+	v := strings.TrimSpace(os.Getenv(key))
+	if v == "" {
+		return defaultValue, nil
+	}
+
+	addr, err := netip.ParseAddr(v)
+	if err != nil {
+		return "", fmt.Errorf(
+			"%w: %s: %q must be an IP address literal such as "+
+				"127.0.0.1, 0.0.0.0 or ::, not a hostname and not "+
+				"host:port: %w",
+			ErrInvalidBindAddress, key, v, err,
+		)
+	}
+
+	return addr.String(), nil
+}
+
+// envSentryDSN returns the value of the named environment variable
+// checked as a Sentry DSN. An unset (or empty, or whitespace-only)
+// value yields "", which means error reporting stays off — the common
+// case, and a normal start.
+//
+// A set value is parsed with sentry.NewDsn, which is the call
+// sentry.Init makes on the DSN it is handed, so what passes here is
+// exactly what the SDK will accept later and the two cannot disagree.
+// Reproducing the check by hand instead would cost this package its
+// dependency on the SDK — already a module dependency, already linked
+// into the binary — in exchange for a second definition of "valid DSN"
+// free to drift from the one that decides.
+//
+// A set value that does not parse is a hard error naming the key, so
+// startup fails loudly. Losing error reporting is the failure this
+// variable exists to prevent, and a typo in a DSN is silent forever:
+// nothing later in the process can notice that reports are going
+// nowhere. The bad value is quoted because it is a URL to a public
+// endpoint carrying a public key, not a secret.
+func envSentryDSN(key string) (string, error) {
+	v := strings.TrimSpace(os.Getenv(key))
+	if v == "" {
+		return "", nil
+	}
+
+	_, err := sentry.NewDsn(v)
+	if err != nil {
+		return "", fmt.Errorf(
+			"%w: %s: %q: %w", ErrInvalidSentryDSN, key, v, err,
+		)
+	}
+
+	return v, nil
+}
+
 // resolveMetricsAuth reads the /metrics basic-auth credentials and
 // rejects a half-set pair, naming both variables either way. The
 // error carries neither value: the password is a secret.
@@ -411,12 +586,14 @@ func resolveMetricsAuth() (string, string, error) {
 	)
 }

-// resolveEnvironment reads WEBHOOKER_ENVIRONMENT, defaulting to
-// dev, and rejects unrecognised values.
+// resolveEnvironment reads WEBHOOKER_ENVIRONMENT, defaulting to prod
+// when it is unset so a deployment that forgets the variable is not
+// silently permissive; dev must be set explicitly. It rejects
+// unrecognised values.
 func resolveEnvironment() (string, error) {
 	environment := os.Getenv("WEBHOOKER_ENVIRONMENT")
 	if environment == "" {
-		environment = EnvironmentDev
+		environment = EnvironmentProd
 	}

 	if environment != EnvironmentDev &&
@@ -431,6 +608,27 @@ func resolveEnvironment() (string, error) {
 	return environment, nil
 }

+// resolveListener reads the two variables that describe the HTTP
+// listener: which port it claims and which address it claims it on.
+// They are read together because neither is meaningful alone, and
+// because a validation failure in either has to abort startup before
+// anything binds.
+func resolveListener() (int, string, error) {
+	port, err := envPort("PORT", defaultPort)
+	if err != nil {
+		return 0, "", err
+	}
+
+	bindAddress, err := envBindAddress(
+		"BIND_ADDRESS", defaultBindAddress,
+	)
+	if err != nil {
+		return 0, "", err
+	}
+
+	return port, bindAddress, nil
+}
+
 // loadFromEnv builds a Config from the environment. Every value that
 // needs parsing fails loudly when it is set but unparseable: the
 // documented defaults apply only to variables that are unset (or
@@ -442,7 +640,7 @@ func loadFromEnv() (*Config, error) {
 		return nil, err
 	}

-	port, err := envPort("PORT", defaultPort)
+	port, bindAddress, err := resolveListener()
 	if err != nil {
 		return nil, err
 	}
@@ -498,6 +696,11 @@ func loadFromEnv() (*Config, error) {
 		return nil, err
 	}

+	sentryDSN, err := envSentryDSN("SENTRY_DSN")
+	if err != nil {
+		return nil, err
+	}
+
 	return &Config{
 		DataDir:                DataDir(),
 		Debug:                  debug,
@@ -506,7 +709,8 @@ func loadFromEnv() (*Config, error) {
 		MetricsUsername:        metricsUsername,
 		MetricsPassword:        metricsPassword,
 		Port:                   port,
-		SentryDSN:              envString("SENTRY_DSN"),
+		BindAddress:            bindAddress,
+		SentryDSN:              sentryDSN,
 		RetentionSweepInterval: retentionSweepInterval,
 		SessionIdleTimeout:     sessionIdleTimeout,
 		ReceiverRateLimit:      receiverRateLimit,
@@ -543,12 +747,14 @@ func (c *Config) warnEgressAllowlist(log *slog.Logger) {

 	log.Warn(
 		"ALLOWED_EGRESS_CIDRS lets delivery targets reach these "+
-			"otherwise-blocked private/reserved networks. Anyone "+
-			"who can create a delivery target can now make this "+
-			"process issue requests into them, and read back the "+
-			"response. Link-local and the known cloud instance "+
-			"metadata endpoints outside it stay blocked "+
-			"regardless of what is listed here.",
+			"otherwise-blocked networks. Anyone who can create a "+
+			"delivery target can now make this process issue "+
+			"requests into them, and read back the response. Only "+
+			"the addresses the README lists as blocked "+
+			"unconditionally stay blocked regardless of what is "+
+			"listed here; a public cloud metadata address such as "+
+			"168.63.129.16 is reachable once it, or a block "+
+			"covering it, is listed.",
 		"allowedEgressCIDRs",
 		strings.Join(PrefixStrings(c.AllowedEgressCIDRs), ","),
 	)
@@ -571,10 +777,8 @@ func (c *Config) warnEgressAllowlist(log *slog.Logger) {
 // everyone else's wrong passwords, and the receiver's limits become
 // service-wide ceilings.
 //
-// The warning is deliberately not gated on WEBHOOKER_ENVIRONMENT. That
-// variable defaults to dev, so gating on it would silence the warning
-// for exactly the operator who forgot to configure the deployment —
-// the case it exists to catch.
+// The warning is deliberately not gated on WEBHOOKER_ENVIRONMENT:
+// behind a proxy every client shares one bucket in dev and prod alike.
 //
 // The default of trusting nobody is deliberate — trusting forwarded
 // headers from arbitrary peers lets any client choose its own bucket —
@@ -625,6 +829,11 @@ func New(lc fx.Lifecycle, params ConfigParams) (*Config, error) {
 	log.Info("Configuration loaded",
 		"environment", s.Environment,
 		"port", s.Port,
+		// Logged because which interfaces the cleartext listener
+		// answers on is not otherwise observable from inside a
+		// container, and it decides whether anything but the local
+		// host can reach the admin UI.
+		"bindAddress", s.BindAddress,
 		"debug", s.Debug,
 		"maintenanceMode", s.MaintenanceMode,
 		"dataDir", s.DataDir,
@@ -636,7 +845,7 @@ func New(lc fx.Lifecycle, params ConfigParams) (*Config, error) {
 		"receiverRateLimit", s.ReceiverRateLimit,
 		"trustedProxies", len(s.TrustedProxies),
 		"allowedEgressCIDRs", len(s.AllowedEgressCIDRs),
-		"hasSentryDSN", s.SentryDSN != "",
+		"sentryEnabled", s.SentryEnabled(),
 		"hasMetricsAuth", s.MetricsAuthEnabled(),
 	)

@@ -44,9 +44,9 @@ func TestEnvironmentConfig(t *testing.T) {
 		isProd      bool
 	}{
 		{
-			name:   "default is dev",
-			isDev:  true,
-			isProd: false,
+			name:   "default is prod",
+			isDev:  false,
+			isProd: true,
 		},
 		{
 			name:     "explicit dev",
@@ -834,12 +834,13 @@ func TestEgressAllowlistWarning(t *testing.T) {
 			// to be able to read back which networks are open.
 			assert.Contains(t, logged, "10.0.0.0/8")
 			assert.Contains(t, logged, "127.0.0.0/8")
-			// What stays shut. Asserted on the clause naming the
-			// wider set rather than on "Link-local" alone, so the
-			// string cannot narrow back to link-local only while
-			// the always-blocked set covers ULA, CGNAT and two
-			// public metadata addresses as well.
-			assert.Contains(t, logged, "metadata endpoints outside it")
+			// What stays shut is the whole unconditional set, not
+			// link-local alone; a public metadata address is not in
+			// it, so a listed block covering it opens it.
+			assert.Contains(t, logged, "blocked unconditionally")
+			assert.Contains(t, logged, "168.63.129.16 is reachable")
+			// The listed blocks need not be private or reserved.
+			assert.NotContains(t, logged, "private/reserved")
 		})
 	}
 }
@@ -848,10 +849,9 @@ func TestEgressAllowlistWarning(t *testing.T) {
 // tells an operator a deployment behind a reverse proxy shares one
 // rate-limit bucket between every client, which turns the receiver
 // limits into service-wide ceilings and collapses login failure
-// counting. It must fire whenever TRUSTED_PROXIES is empty,
-// in any environment: WEBHOOKER_ENVIRONMENT defaults to dev, so gating
-// on it would silence the warning for exactly the operator who never
-// configured the deployment. It stays quiet once proxies are named.
+// counting. It must fire whenever TRUSTED_PROXIES is empty, in any
+// environment, because behind a proxy every client shares one bucket
+// in dev and prod alike. It stays quiet once proxies are named.
 func TestSharedRateLimitBucketWarning(t *testing.T) {
 	tests := []struct {
 		name           string
@@ -871,10 +871,6 @@ func TestSharedRateLimitBucketWarning(t *testing.T) {
 			expectWarning:  false,
 		},
 		{
-			// The default environment. An internet-exposed
-			// deployment whose operator never set
-			// WEBHOOKER_ENVIRONMENT lands here and has exactly
-			// the exposure the warning announces.
 			name:          "dev without trusted proxies warns",
 			environment:   config.EnvironmentDev,
 			expectWarning: true,
@@ -0,0 +1,158 @@
+package config_test
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/webhooker/internal/config"
+)
+
+// dotEnvKey is a throwaway variable name the .env tests write and
+// read, so they cannot disturb real configuration.
+const dotEnvKey = "WEBHOOKER_TEST_DOTENV_VALUE"
+
+// malformedDotEnv is a file godotenv cannot parse. The first line is
+// the realistic typo — a space where the `=` belongs — and the rest
+// make sure nothing downstream treats the file as salvageable line by
+// line.
+const malformedDotEnv = "PORT 19615\n" +
+	"this is not = valid ! syntax\n" +
+	"\"unclosed\n"
+
+// unsetDotEnvKey makes dotEnvKey genuinely absent for the duration of
+// the test and restores it afterwards. t.Setenv registers the restore;
+// the Unsetenv that follows is what the test actually needs, because a
+// variable set to the empty string is still present in os.Environ and
+// godotenv would refuse to overwrite it.
+func unsetDotEnvKey(t *testing.T) {
+	t.Helper()
+	t.Setenv(dotEnvKey, "placeholder")
+	require.NoError(t, os.Unsetenv(dotEnvKey))
+}
+
+// writeDotEnv writes contents to a .env file in a fresh temporary
+// directory and returns its path.
+func writeDotEnv(t *testing.T, contents string) string {
+	t.Helper()
+
+	path := filepath.Join(t.TempDir(), config.DotEnvPath)
+	require.NoError(t, os.WriteFile(path, []byte(contents), 0o600))
+
+	return path
+}
+
+// TestLoadDotEnv_MissingFileIsFine pins the case most deployments are
+// in. The file is optional: it is a development convenience, and a
+// deployment that configures the environment directly must start
+// normally rather than be refused for a file it was never meant to
+// have.
+//
+//nolint:paralleltest // unsetDotEnvKey uses t.Setenv.
+func TestLoadDotEnv_MissingFileIsFine(t *testing.T) {
+	unsetDotEnvKey(t)
+
+	absent := filepath.Join(t.TempDir(), config.DotEnvPath)
+	require.NoError(t, config.LoadDotEnvFileForTest(absent))
+
+	_, present := os.LookupEnv(dotEnvKey)
+	assert.False(t, present, "nothing may be set from an absent file")
+}
+
+// TestLoadDotEnv_AppliesValues pins that a well-formed file still
+// reaches the environment, which is the whole reason the file is read
+// at all.
+//
+//nolint:paralleltest // unsetDotEnvKey uses t.Setenv.
+func TestLoadDotEnv_AppliesValues(t *testing.T) {
+	unsetDotEnvKey(t)
+
+	path := writeDotEnv(t, "# a comment\n"+dotEnvKey+"=from-dot-env\n")
+
+	require.NoError(t, config.LoadDotEnvFileForTest(path))
+	assert.Equal(t, "from-dot-env", os.Getenv(dotEnvKey))
+}
+
+// TestLoadDotEnv_RealEnvironmentWins pins that the file cannot
+// override a variable the process was actually started with. A
+// deployment that sets DATA_DIR in its unit file must not have it
+// silently replaced by a stale .env left in the working directory.
+func TestLoadDotEnv_RealEnvironmentWins(t *testing.T) {
+	t.Setenv(dotEnvKey, "from-environment")
+
+	path := writeDotEnv(t, dotEnvKey+"=from-dot-env\n")
+
+	require.NoError(t, config.LoadDotEnvFileForTest(path))
+	assert.Equal(t, "from-environment", os.Getenv(dotEnvKey))
+}
+
+// TestLoadDotEnv_MalformedFileAborts is the defect this fixes. One bad
+// line makes godotenv apply none of the file, so every variable in it
+// reverts to its default; the process used to start that way with no
+// log line naming the file at all.
+//
+//nolint:paralleltest // unsetDotEnvKey uses t.Setenv.
+func TestLoadDotEnv_MalformedFileAborts(t *testing.T) {
+	unsetDotEnvKey(t)
+
+	path := writeDotEnv(
+		t, malformedDotEnv+dotEnvKey+"=from-dot-env\n",
+	)
+
+	err := config.LoadDotEnvFileForTest(path)
+
+	require.Error(t, err)
+	require.ErrorIs(t, err, config.ErrDotEnvUnreadable)
+	assert.Contains(
+		t, err.Error(), config.DotEnvPath,
+		"the failure must name the file it could not read",
+	)
+
+	_, present := os.LookupEnv(dotEnvKey)
+	assert.False(
+		t, present,
+		"a rejected file must apply nothing, not part of itself",
+	)
+}
+
+// TestLoadDotEnv_UnreadableFileAborts pins that only absence is
+// tolerated. A .env that exists but cannot be read is a file the
+// operator meant to be applied, so it fails like a malformed one
+// rather than being treated as though it were not there.
+func TestLoadDotEnv_UnreadableFileAborts(t *testing.T) {
+	t.Parallel()
+
+	// A directory in the file's place: open succeeds and the read
+	// fails, which no umask or root-ness can turn back into success
+	// the way a chmod could.
+	path := filepath.Join(t.TempDir(), config.DotEnvPath)
+	require.NoError(t, os.Mkdir(path, 0o750))
+
+	err := config.LoadDotEnvFileForTest(path)
+
+	require.Error(t, err)
+	require.ErrorIs(t, err, config.ErrDotEnvUnreadable)
+}
+
+// TestLoadDotEnv_ReadsTheWorkingDirectory pins the path LoadDotEnv
+// itself opens, which the tests above bypass. It is relative to the
+// process working directory, as it was under godotenv/autoload and as
+// the README documents.
+//
+//nolint:paralleltest // t.Chdir moves the whole process.
+func TestLoadDotEnv_ReadsTheWorkingDirectory(t *testing.T) {
+	unsetDotEnvKey(t)
+
+	dir := t.TempDir()
+	require.NoError(t, os.WriteFile(
+		filepath.Join(dir, config.DotEnvPath),
+		[]byte(dotEnvKey+"=from-working-directory\n"),
+		0o600,
+	))
+	t.Chdir(dir)
+
+	require.NoError(t, config.LoadDotEnv())
+	assert.Equal(t, "from-working-directory", os.Getenv(dotEnvKey))
+}
@@ -21,6 +21,22 @@ const (
 	envKeyPort            = "PORT"
 	envKeyDebug           = "DEBUG"
 	envKeyMaintenanceMode = "MAINTENANCE_MODE"
+	envKeyBindAddress     = "BIND_ADDRESS"
+)
+
+// Sample BIND_ADDRESS values used by the tables below.
+const (
+	// bindAddressDefault is the shipped default. It is asserted
+	// against the package's own constant in
+	// TestNewUsesDefaultsWhenUnset, so the two cannot drift.
+	bindAddressDefault = "127.0.0.1"
+
+	// bindAddressWildcard is the value a container deployment sets.
+	bindAddressWildcard = "0.0.0.0"
+
+	// bindAddressSample is an arbitrary specific address, standing
+	// for "one interface of several".
+	bindAddressSample = "10.1.2.3"
 )

 // envBoolCase is one row of the envBool table.
@@ -291,6 +307,160 @@ func TestEnvPort(t *testing.T) {
 	}
 }

+// TestEnvBindAddress covers BIND_ADDRESS parsing.
+//
+// Only IP address literals are accepted. Every rejection below is a
+// value an operator plausibly writes — a hostname, a host:port, a
+// CIDR block — and each has to abort startup rather than fall back to
+// the default, because falling back would bind an address other than
+// the one asked for and, in the wildcard-default case this setting
+// exists to end, publish cleartext on every interface.
+func TestEnvBindAddress(t *testing.T) {
+	for _, tt := range envBindAddressCases() {
+		t.Run(tt.name, func(t *testing.T) {
+			// Cannot use t.Parallel() here because t.Setenv
+			// is incompatible with parallel subtests.
+			if tt.set {
+				t.Setenv(testEnvKey, tt.value)
+			} else {
+				require.NoError(t, os.Unsetenv(testEnvKey))
+			}
+
+			got, err := config.EnvBindAddressForTest(
+				testEnvKey, bindAddressDefault,
+			)
+
+			if tt.expectError {
+				require.Error(t, err)
+				require.ErrorIs(t, err, config.ErrInvalidBindAddress)
+				assert.Contains(t, err.Error(), testEnvKey)
+				assert.Contains(t, err.Error(), tt.value)
+
+				return
+			}
+
+			require.NoError(t, err)
+			assert.Equal(t, tt.expected, got)
+		})
+	}
+}
+
+// envBindAddressCase is one row of the envBindAddress table.
+type envBindAddressCase struct {
+	name        string
+	set         bool
+	value       string
+	expectError bool
+	expected    string
+}
+
+// envBindAddressCases is the envBindAddress table, kept out of the
+// test body so the test itself stays readable.
+func envBindAddressCases() []envBindAddressCase {
+	return append(
+		envBindAddressAcceptedCases(),
+		envBindAddressRejectedCases()...,
+	)
+}
+
+// envBindAddressAcceptedCases are the values that parse: the three
+// spellings of "unset" that take the default, and the literals.
+func envBindAddressAcceptedCases() []envBindAddressCase {
+	return []envBindAddressCase{
+		{
+			name:     "unset returns the default",
+			expected: bindAddressDefault,
+		},
+		{
+			name:     "empty returns the default",
+			set:      true,
+			value:    "",
+			expected: bindAddressDefault,
+		},
+		{
+			name:     "whitespace returns the default",
+			set:      true,
+			value:    "   ",
+			expected: bindAddressDefault,
+		},
+		{
+			name:     "ipv4 wildcard is parsed",
+			set:      true,
+			value:    bindAddressWildcard,
+			expected: bindAddressWildcard,
+		},
+		{
+			name:     "ipv4 literal is parsed",
+			set:      true,
+			value:    bindAddressSample,
+			expected: bindAddressSample,
+		},
+		{
+			name:     "surrounding whitespace is trimmed",
+			set:      true,
+			value:    "  " + bindAddressSample + "  ",
+			expected: bindAddressSample,
+		},
+		{
+			name:     "ipv6 wildcard is parsed",
+			set:      true,
+			value:    "::",
+			expected: "::",
+		},
+		{
+			name:     "ipv6 literal is parsed",
+			set:      true,
+			value:    "2001:db8::5",
+			expected: "2001:db8::5",
+		},
+	}
+}
+
+// envBindAddressRejectedCases are the values that abort startup.
+// Each is something an operator plausibly writes, and none may fall
+// back to the default: the default is loopback, so a silent fallback
+// would bind somewhere other than what was asked for.
+func envBindAddressRejectedCases() []envBindAddressCase {
+	return []envBindAddressCase{
+		{
+			name:        "garbage is rejected",
+			set:         true,
+			value:       "not-an-address",
+			expectError: true,
+		},
+		{
+			name:        "hostname is rejected",
+			set:         true,
+			value:       "localhost",
+			expectError: true,
+		},
+		{
+			name:        "unresolvable hostname is rejected",
+			set:         true,
+			value:       "no-such-host.invalid",
+			expectError: true,
+		},
+		{
+			name:        "host and port is rejected",
+			set:         true,
+			value:       bindAddressDefault + ":8080",
+			expectError: true,
+		},
+		{
+			name:        "bracketed ipv6 is rejected",
+			set:         true,
+			value:       "[::1]",
+			expectError: true,
+		},
+		{
+			name:        "CIDR block is rejected",
+			set:         true,
+			value:       "10.0.0.0/8",
+			expectError: true,
+		},
+	}
+}
+
 // buildConfig constructs a Config through fx exactly as the
 // application does, returning the config and any construction error.
 func buildConfig(t *testing.T) (*config.Config, error) {
@@ -312,58 +482,7 @@ func buildConfig(t *testing.T) (*config.Config, error) {
 }

 func TestNewRejectsBadEnvValues(t *testing.T) {
-	tests := []struct {
-		name        string
-		key         string
-		value       string
-		expectError bool
-		check       func(t *testing.T, cfg *config.Config)
-	}{
-		{
-			name:  "valid PORT is used",
-			key:   envKeyPort,
-			value: "9001",
-			check: func(t *testing.T, cfg *config.Config) {
-				t.Helper()
-				assert.Equal(t, 9001, cfg.Port)
-			},
-		},
-		{
-			name:        "unparseable PORT aborts startup",
-			key:         envKeyPort,
-			value:       "eighty-eighty",
-			expectError: true,
-		},
-		{
-			name:        "out-of-range PORT aborts startup",
-			key:         envKeyPort,
-			value:       "70000",
-			expectError: true,
-		},
-		{
-			name:  "valid DEBUG is used",
-			key:   envKeyDebug,
-			value: "true",
-			check: func(t *testing.T, cfg *config.Config) {
-				t.Helper()
-				assert.True(t, cfg.Debug)
-			},
-		},
-		{
-			name:        "unparseable DEBUG aborts startup",
-			key:         envKeyDebug,
-			value:       "ture",
-			expectError: true,
-		},
-		{
-			name:        "unparseable MAINTENANCE_MODE aborts startup",
-			key:         envKeyMaintenanceMode,
-			value:       "sometimes",
-			expectError: true,
-		},
-	}
-
-	for _, tt := range tests {
+	for _, tt := range badEnvValueCases() {
 		t.Run(tt.name, func(t *testing.T) {
 			// Cannot use t.Parallel() here because t.Setenv
 			// is incompatible with parallel subtests.
@@ -387,6 +506,149 @@ func TestNewRejectsBadEnvValues(t *testing.T) {
 	}
 }

+// badEnvValueCase is one row of the config.New table: a variable, the
+// value it is set to, and either the assertion that startup fails
+// naming both, or a check on the Config that resulted.
+type badEnvValueCase struct {
+	name        string
+	key         string
+	value       string
+	expectError bool
+	check       func(t *testing.T, cfg *config.Config)
+}
+
+// badEnvValueCases is the config.New table, kept out of the test body
+// so the test itself stays readable. It is assembled from per-variable
+// groups because one literal covering every variable outgrew the
+// function-length budget.
+func badEnvValueCases() []badEnvValueCase {
+	cases := listenerEnvValueCases()
+	cases = append(cases, flagEnvValueCases()...)
+	cases = append(cases, sentryEnvValueCases()...)
+
+	return cases
+}
+
+// listenerEnvValueCases covers the two variables that describe the
+// HTTP listener.
+func listenerEnvValueCases() []badEnvValueCase {
+	return []badEnvValueCase{
+		{
+			name:  "valid PORT is used",
+			key:   envKeyPort,
+			value: "9001",
+			check: func(t *testing.T, cfg *config.Config) {
+				t.Helper()
+				assert.Equal(t, 9001, cfg.Port)
+			},
+		},
+		{
+			name:        "unparseable PORT aborts startup",
+			key:         envKeyPort,
+			value:       "eighty-eighty",
+			expectError: true,
+		},
+		{
+			name:        "out-of-range PORT aborts startup",
+			key:         envKeyPort,
+			value:       "70000",
+			expectError: true,
+		},
+		{
+			name:  "valid BIND_ADDRESS is used",
+			key:   envKeyBindAddress,
+			value: bindAddressWildcard,
+			check: func(t *testing.T, cfg *config.Config) {
+				t.Helper()
+				assert.Equal(
+					t, bindAddressWildcard, cfg.BindAddress,
+				)
+			},
+		},
+		{
+			name:        "unparseable BIND_ADDRESS aborts startup",
+			key:         envKeyBindAddress,
+			value:       "not-an-address",
+			expectError: true,
+		},
+		{
+			name:        "hostname BIND_ADDRESS aborts startup",
+			key:         envKeyBindAddress,
+			value:       "localhost",
+			expectError: true,
+		},
+		{
+			name:        "BIND_ADDRESS with a port aborts startup",
+			key:         envKeyBindAddress,
+			value:       bindAddressDefault + ":8080",
+			expectError: true,
+		},
+	}
+}
+
+// flagEnvValueCases covers the boolean variables.
+func flagEnvValueCases() []badEnvValueCase {
+	return []badEnvValueCase{
+		{
+			name:  "valid DEBUG is used",
+			key:   envKeyDebug,
+			value: "true",
+			check: func(t *testing.T, cfg *config.Config) {
+				t.Helper()
+				assert.True(t, cfg.Debug)
+			},
+		},
+		{
+			name:        "unparseable DEBUG aborts startup",
+			key:         envKeyDebug,
+			value:       "ture",
+			expectError: true,
+		},
+		{
+			name:        "unparseable MAINTENANCE_MODE aborts startup",
+			key:         envKeyMaintenanceMode,
+			value:       "sometimes",
+			expectError: true,
+		},
+	}
+}
+
+// sentryEnvValueCases covers SENTRY_DSN. The three rejected values are
+// the ones measured on the defect: each initialised the SDK with an
+// error and left the process serving with error reporting off.
+func sentryEnvValueCases() []badEnvValueCase {
+	return []badEnvValueCase{
+		{
+			name:  "valid SENTRY_DSN is used",
+			key:   envKeySentryDSN,
+			value: validSentryDSN,
+			check: func(t *testing.T, cfg *config.Config) {
+				t.Helper()
+				assert.Equal(t, validSentryDSN, cfg.SentryDSN)
+				assert.True(t, cfg.SentryEnabled())
+			},
+		},
+		{
+			name:        "unparseable SENTRY_DSN aborts startup",
+			key:         envKeySentryDSN,
+			value:       "not-a-dsn",
+			expectError: true,
+		},
+		{
+			name:        "SENTRY_DSN that is not a URL aborts startup",
+			key:         envKeySentryDSN,
+			value:       "%%%",
+			expectError: true,
+		},
+		{
+			name:        "keyless SENTRY_DSN aborts startup",
+			key:         envKeySentryDSN,
+			value:       "https://example.invalid/1",
+			expectError: true,
+		},
+	}
+}
+
 // TestNewUsesDefaultsWhenUnset proves the fail-loud behaviour did not
 // break the legitimate unset case: absent variables still get their
 // documented defaults.
@@ -395,6 +657,7 @@ func TestNewUsesDefaultsWhenUnset(t *testing.T) {

 	for _, key := range []string{
 		envKeyPort, envKeyDebug, envKeyMaintenanceMode,
+		envKeyBindAddress, envKeySentryDSN,
 	} {
 		require.NoError(t, os.Unsetenv(key))
 	}
@@ -406,4 +669,20 @@ func TestNewUsesDefaultsWhenUnset(t *testing.T) {
 	assert.Equal(t, 8080, cfg.Port)
 	assert.False(t, cfg.Debug)
 	assert.False(t, cfg.MaintenanceMode)
+
+	// Loopback, not the wildcard: the default must not publish the
+	// cleartext admin UI and the unauthenticated receiver on every
+	// interface of a host that configured nothing. The value is read
+	// from the package rather than repeated, so the README's
+	// documented default and the compiled-in one are pinned to the
+	// same constant.
+	assert.Equal(
+		t, config.DefaultBindAddressForTest, cfg.BindAddress,
+	)
+	assert.Equal(t, bindAddressDefault, cfg.BindAddress)
+
+	// An absent SENTRY_DSN is the common case and must stay a normal
+	// start with error reporting off, not a refusal.
+	assert.Empty(t, cfg.SentryDSN)
+	assert.False(t, cfg.SentryEnabled())
 }
@@ -50,3 +50,25 @@ func EnvPositiveIntForTest(key string, defaultValue int) (int, error) {
 func EnvPortForTest(key string, defaultValue int) (int, error) {
 	return envPort(key, defaultValue)
 }
+
+// EnvSentryDSNForTest exposes envSentryDSN.
+func EnvSentryDSNForTest(key string) (string, error) {
+	return envSentryDSN(key)
+}
+
+// LoadDotEnvFileForTest exposes the loader LoadDotEnv runs, over a
+// caller-named file rather than the process working directory, so
+// each .env state can be covered without moving the test process.
+func LoadDotEnvFileForTest(path string) error {
+	return loadDotEnvFile(path)
+}
+
+// EnvBindAddressForTest exposes envBindAddress.
+func EnvBindAddressForTest(key, defaultValue string) (string, error) {
+	return envBindAddress(key, defaultValue)
+}
+
+// DefaultBindAddressForTest exposes the compiled-in BIND_ADDRESS
+// default, so a test pins the documented value rather than repeating
+// a literal that could drift from it.
+const DefaultBindAddressForTest = defaultBindAddress
@@ -0,0 +1,141 @@
+package config_test
+
+import (
+	"os"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/webhooker/internal/config"
+)
+
+// envKeySentryDSN is the variable envSentryDSN reads in production.
+const envKeySentryDSN = "SENTRY_DSN"
+
+// validSentryDSN is a syntactically complete DSN. The host is under
+// .invalid (RFC 2606), so nothing a test builds around it can reach a
+// real Sentry installation.
+const validSentryDSN = "https://abc123@sentry.invalid/42"
+
+// envSentryDSNCase is one row of the envSentryDSN table.
+type envSentryDSNCase struct {
+	name        string
+	set         bool
+	value       string
+	expectError bool
+	expected    string
+}
+
+// envSentryDSNCases is the envSentryDSN table. The three invalid
+// values are the ones measured on the defect: each initialised the SDK
+// with an error and left the process serving with reporting off.
+func envSentryDSNCases() []envSentryDSNCase {
+	return []envSentryDSNCase{
+		{
+			name:     "unset means reporting off",
+			expected: "",
+		},
+		{
+			name:     "empty means reporting off",
+			set:      true,
+			value:    "",
+			expected: "",
+		},
+		{
+			name:     "whitespace means reporting off",
+			set:      true,
+			value:    "   ",
+			expected: "",
+		},
+		{
+			name:     "a valid DSN is kept",
+			set:      true,
+			value:    validSentryDSN,
+			expected: validSentryDSN,
+		},
+		{
+			name:     "surrounding whitespace is trimmed",
+			set:      true,
+			value:    "  " + validSentryDSN + "\t",
+			expected: validSentryDSN,
+		},
+		{
+			name:        "a value that is not a URL is rejected",
+			set:         true,
+			value:       "not-a-dsn",
+			expectError: true,
+		},
+		{
+			name:        "an unparseable URL is rejected",
+			set:         true,
+			value:       "%%%",
+			expectError: true,
+		},
+		{
+			name:        "a DSN without a public key is rejected",
+			set:         true,
+			value:       "https://example.invalid/1",
+			expectError: true,
+		},
+		{
+			name:        "a DSN without a project id is rejected",
+			set:         true,
+			value:       "https://abc123@sentry.invalid/",
+			expectError: true,
+		},
+		{
+			name:        "a non-HTTP scheme is rejected",
+			set:         true,
+			value:       "ftp://abc123@sentry.invalid/42",
+			expectError: true,
+		},
+	}
+}
+
+// TestEnvSentryDSN covers the helper directly. What it pins beyond the
+// value is the failure shape: a set-but-unparseable DSN names the
+// variable and the value, exactly as the other fail-loud helpers do,
+// so an operator reads the fix off the message.
+func TestEnvSentryDSN(t *testing.T) {
+	for _, tt := range envSentryDSNCases() {
+		t.Run(tt.name, func(t *testing.T) {
+			// Cannot use t.Parallel() here because t.Setenv
+			// is incompatible with parallel subtests.
+			if tt.set {
+				t.Setenv(envKeySentryDSN, tt.value)
+			} else {
+				require.NoError(t, os.Unsetenv(envKeySentryDSN))
+			}
+
+			got, err := config.EnvSentryDSNForTest(envKeySentryDSN)
+
+			if tt.expectError {
+				require.Error(t, err)
+				require.ErrorIs(t, err, config.ErrInvalidSentryDSN)
+				assert.Contains(t, err.Error(), envKeySentryDSN)
+				assert.Contains(t, err.Error(), tt.value)
+				assert.Empty(t, got)
+
+				return
+			}
+
+			require.NoError(t, err)
+			assert.Equal(t, tt.expected, got)
+		})
+	}
+}
+
+// TestSentryEnabled_TracksTheDSN pins that the one method answering
+// "is anything being reported" agrees with the DSN in every state. The
+// startup log, the SDK initialisation and the sentryhttp middleware
+// all read it, so a log field cannot report reporting as on while
+// nothing is sending.
+func TestSentryEnabled_TracksTheDSN(t *testing.T) {
+	t.Parallel()
+
+	assert.False(t, (&config.Config{}).SentryEnabled())
+	assert.True(
+		t,
+		(&config.Config{SentryDSN: validSentryDSN}).SentryEnabled(),
+	)
+}
@@ -3,6 +3,8 @@ package database_test
 import (
 	"bytes"
 	"context"
+	"log/slog"
+	"path/filepath"
 	"strings"
 	"testing"

@@ -83,3 +85,37 @@ func TestFirstBoot_PrintsTheAdminPasswordAsABanner(t *testing.T) {
 		t, ok, "the printed password must open the seeded account",
 	)
 }
+
+// TestNewDatabase_IsLoggedWithItsPath is the log half of
+// https://git.eeqj.de/sneak/webhooker/issues/359. A DATA_DIR that is
+// unexpectedly empty boots exactly like a first start, so the start
+// that creates the database must say so, and where. Opening that
+// database again must not.
+func TestNewDatabase_IsLoggedWithItsPath(t *testing.T) {
+	t.Parallel()
+
+	dir := t.TempDir()
+
+	open := func() string {
+		var out bytes.Buffer
+
+		db, err := database.Open(dir, slog.New(slog.NewTextHandler(&out, nil)))
+		require.NoError(t, err)
+		require.NoError(t, db.Close())
+
+		return out.String()
+	}
+
+	const created = `level=WARN msg="created a new, empty database"`
+
+	first := open()
+	second := open()
+
+	assert.Contains(
+		t, first,
+		created+" path="+filepath.Join(dir, database.MainDBFileName),
+	)
+	assert.NotContains(
+		t, second, created, "an existing database is not new",
+	)
+}
@@ -4,11 +4,11 @@ package database
 import (
 	"context"
 	"crypto/rand"
-	"database/sql"
 	"encoding/base64"
 	"errors"
 	"fmt"
 	"io"
+	"io/fs"
 	"log/slog"
 	"os"
 	"path/filepath"
@@ -16,15 +16,14 @@ import (
 	"go.uber.org/fx"
 	"gorm.io/driver/sqlite"
 	"gorm.io/gorm"
-	_ "modernc.org/sqlite" // Pure Go SQLite driver
 	"sneak.berlin/go/webhooker/internal/banner"
 	"sneak.berlin/go/webhooker/internal/config"
+	"sneak.berlin/go/webhooker/internal/datadir"
 	"sneak.berlin/go/webhooker/internal/gormlog"
 	"sneak.berlin/go/webhooker/internal/logger"
 )

 const (
-	dataDirPerm       = 0750
 	randomPasswordLen = 16
 	sessionKeyLen     = 32
 )
@@ -187,7 +186,9 @@ func (d *Database) connect() error {
 // caller's decision.
 func (d *Database) connectTo(dataDir string) error {
 	// Ensure the data directory exists before opening the database.
-	err := os.MkdirAll(dataDir, dataDirPerm)
+	// datadir.DirPerm is the single source of the directory mode; this
+	// package creates the directory too, since either may run first.
+	err := os.MkdirAll(dataDir, datadir.DirPerm)
 	if err != nil {
 		return fmt.Errorf(
 			"creating data directory %s: %w",
@@ -198,13 +199,17 @@ func (d *Database) connectTo(dataDir string) error {

 	// Construct the main application database path inside DATA_DIR.
 	dbPath := filepath.Join(dataDir, MainDBFileName)
-	dbURL := fmt.Sprintf(
-		"file:%s?cache=shared&mode=rwc",
-		dbPath,
-	)

-	// Open the database with the pure Go SQLite driver
-	sqlDB, err := sql.Open("sqlite", dbURL)
+	// Checked before opening, which creates the file. A DATA_DIR that
+	// is unexpectedly empty -- its volume not mounted, say -- looks
+	// exactly like a first start, so a new database is a warning.
+	_, statErr := os.Stat(dbPath)
+	created := errors.Is(statErr, fs.ErrNotExist)
+
+	// Opened through OpenSQLite so this handle carries the same WAL
+	// journaling, busy timeout, immediate-transaction locking, and pool
+	// bounds as every other database file. See sqlite_open.go.
+	sqlDB, err := OpenSQLite(dbPath, SQLiteModeCreate)
 	if err != nil {
 		d.log.Error(
 			"failed to open database",
@@ -231,7 +236,12 @@ func (d *Database) connectTo(dataDir string) error {
 	}

 	d.db = db
-	d.log.Info("connected to database", "path", dbPath)
+
+	if created {
+		d.log.Warn("created a new, empty database", "path", dbPath)
+	} else {
+		d.log.Info("connected to database", "path", dbPath)
+	}

 	// Run migrations
 	return d.migrate()
@@ -0,0 +1,166 @@
+package database_test
+
+import (
+	"context"
+	"fmt"
+	"testing"
+	"time"
+
+	"github.com/google/uuid"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"gorm.io/gorm"
+	"sneak.berlin/go/webhooker/internal/database"
+)
+
+// TestWebhookDBManager_OpenAddsEventTierIndexes verifies that opening a
+// per-webhook database that predates these indexes creates them. It
+// stands in for an older database file by dropping the indexes
+// AutoMigrate just created, then reopening the same file.
+func TestWebhookDBManager_OpenAddsEventTierIndexes(t *testing.T) {
+	t.Parallel()
+
+	indexes := []struct {
+		model any
+		name  string
+	}{
+		{&database.Delivery{}, "idx_deliveries_status"},
+		{&database.Delivery{}, "idx_deliveries_event_id"},
+		{&database.DeliveryResult{}, "idx_delivery_results_delivery_id"},
+		{&database.Event{}, "idx_events_deleted_at_created_at"},
+		{&database.Event{}, "idx_events_created_at"},
+	}
+
+	mgr, lc := setupTestWebhookDBManager(t)
+	ctx := context.Background()
+	require.NoError(t, lc.Start(ctx))
+
+	defer func() { require.NoError(t, lc.Stop(ctx)) }()
+
+	webhookID := uuid.New().String()
+
+	db, err := mgr.GetDB(webhookID)
+	require.NoError(t, err)
+
+	// A fresh database has them.
+	for _, ix := range indexes {
+		require.True(t, db.Migrator().HasIndex(ix.model, ix.name))
+	}
+
+	// Stand in for a database file created before the indexes existed.
+	for _, ix := range indexes {
+		require.NoError(t, db.Migrator().DropIndex(ix.model, ix.name))
+		require.False(t, db.Migrator().HasIndex(ix.model, ix.name))
+	}
+
+	// Drop the cached connection so the next open reopens the file and
+	// runs AutoMigrate against it, as a restart would.
+	require.NoError(t, mgr.CloseAll())
+
+	db, err = mgr.GetDB(webhookID)
+	require.NoError(t, err)
+
+	for _, ix := range indexes {
+		assert.True(t, db.Migrator().HasIndex(ix.model, ix.name),
+			"opening the existing database should create %s", ix.name)
+	}
+}
+
+// TestEventTierQueriesUseTheirIndexes verifies that the statements the
+// indexes are for use them. GORM builds each statement in a dry run as
+// the code named above it does, soft-delete condition included, and
+// SQLite, which keeps no statistics on these tables, must plan to seek
+// on each index listed by the columns in parentheses.
+func TestEventTierQueriesUseTheirIndexes(t *testing.T) {
+	t.Parallel()
+
+	mgr, lc := setupTestWebhookDBManager(t)
+	ctx := context.Background()
+	require.NoError(t, lc.Start(ctx))
+
+	defer func() { require.NoError(t, lc.Stop(ctx)) }()
+
+	db, err := mgr.GetDB(uuid.New().String())
+	require.NoError(t, err)
+
+	dry := db.Session(&gorm.Session{DryRun: true})
+	ids := []string{
+		uuid.New().String(), uuid.New().String(), uuid.New().String(),
+	}
+	cutoff := time.Now()
+
+	var (
+		deliveries []database.Delivery
+		results    []database.DeliveryResult
+		depths     []struct{ Depth int }
+	)
+
+	byStatus := "idx_deliveries_status (status=? AND deleted_at=?)"
+	byEvent := "idx_deliveries_event_id (event_id=? AND deleted_at=?)"
+	byAge := "idx_events_deleted_at_created_at (deleted_at=? AND created_at<?)"
+
+	// The delivery engine: recovery and the retry sweep, the sweep for
+	// stranded pending deliveries, and the queue depth count.
+	assertPlanUses(t, db, dry.Where(
+		"status = ?", database.DeliveryStatusRetrying,
+	).Find(&deliveries), byStatus)
+	assertPlanUses(t, db, dry.Where(
+		"status = ? AND updated_at < ?",
+		database.DeliveryStatusPending, cutoff,
+	).Limit(500).Find(&deliveries), byStatus)
+	assertPlanUses(t, db, dry.Model(&database.Delivery{}).
+		Select("target_id", "status", "count(*) as depth").
+		Where("status IN ?", []database.DeliveryStatus{
+			database.DeliveryStatusPending,
+			database.DeliveryStatusRetrying,
+		}).Group("target_id, status").Find(&depths), byStatus)
+
+	// The event log: each event's deliveries, then their attempts
+	// (loadEventsWithDeliveries, loadDeliveryResults).
+	assertPlanUses(t, db, dry.Where("event_id = ?", ids[0]).
+		Find(&deliveries), byEvent)
+	assertPlanUses(t, db, dry.Where("delivery_id IN ?", ids).
+		Order("attempt_num ASC").Find(&results),
+		"idx_delivery_results_delivery_id (delivery_id=? AND deleted_at=?)")
+
+	// Retention's three deletes (reapExpired), whose subqueries are built
+	// afresh for each statement as it builds them.
+	expiredEventIDs := func() *gorm.DB {
+		return dry.Model(&database.Event{}).Select("id").
+			Where("created_at < ?", cutoff)
+	}
+
+	assertPlanUses(t, db, dry.Unscoped().Where(
+		"delivery_id IN (?)", dry.Model(&database.Delivery{}).
+			Select("id").Where("event_id IN (?)", expiredEventIDs()),
+	).Delete(&database.DeliveryResult{}),
+		"idx_delivery_results_delivery_id (delivery_id=?)", byEvent, byAge)
+	assertPlanUses(t, db, dry.Unscoped().Where(
+		"event_id IN (?)", expiredEventIDs(),
+	).Delete(&database.Delivery{}),
+		"idx_deliveries_event_id (event_id=?)", byAge)
+	assertPlanUses(t, db, dry.Unscoped().Where(
+		"created_at < ?", cutoff,
+	).Delete(&database.Event{}), "idx_events_created_at (created_at<?)")
+}
+
+// assertPlanUses asserts that SQLite's plan for a statement GORM built
+// in a dry run, run with the same SQL and arguments GORM would send,
+// names each of the given indexes.
+func assertPlanUses(
+	t *testing.T, db, built *gorm.DB, indexes ...string,
+) {
+	t.Helper()
+
+	var plan []struct{ Detail string }
+
+	require.NoError(t, db.Raw(
+		"EXPLAIN QUERY PLAN "+built.Statement.SQL.String(),
+		built.Statement.Vars...,
+	).Scan(&plan).Error)
+
+	for _, index := range indexes {
+		assert.Contains(t, fmt.Sprint(plan), index,
+			built.Statement.SQL.String())
+	}
+}
@@ -1,5 +1,7 @@
 package database

+import "gorm.io/gorm"
+
 // DeliveryStatus represents the status of a delivery
 type DeliveryStatus string

@@ -29,12 +31,19 @@ func (s DeliveryStatus) Terminal() bool {
 }

 // Delivery represents a delivery attempt for an event to a target
+//
+//nolint:lll // a struct tag cannot wrap
 type Delivery struct {
 	BaseModel

-	EventID  string         `gorm:"type:uuid;not null"         json:"eventId"`
-	TargetID string         `gorm:"type:uuid;not null"         json:"targetId"`
-	Status   DeliveryStatus `gorm:"not null;default:'pending'" json:"status"`
+	EventID  string         `gorm:"type:uuid;not null;index:idx_deliveries_event_id,priority:1"       json:"eventId"`
+	TargetID string         `gorm:"type:uuid;not null"                                                json:"targetId"`
+	Status   DeliveryStatus `gorm:"not null;default:'pending';index:idx_deliveries_status,priority:1" json:"status"`
+
+	// DeletedAt repeats the BaseModel field only to be the second column
+	// of the event_id and status indexes, for the reason DeliveryResult
+	// gives.
+	DeletedAt gorm.DeletedAt `gorm:"index:idx_deliveries_event_id,priority:2;index:idx_deliveries_status,priority:2" json:"deletedAt,omitzero"`

 	// Relations
 	Event           Event            `json:"event,omitzero"`
@@ -1,10 +1,21 @@
 package database

+import "gorm.io/gorm"
+
 // DeliveryResult represents the result of a delivery attempt
+//
+//nolint:lll // a struct tag cannot wrap
 type DeliveryResult struct {
 	BaseModel

-	DeliveryID   string `gorm:"type:uuid;not null"   json:"deliveryId"`
+	// DeliveryID and DeletedAt make up one index, in that order.
+	// DeletedAt repeats the BaseModel field only to join it: GORM adds
+	// "deleted_at IS NULL" to almost every query, and where a column is
+	// matched against several values SQLite otherwise reads through the
+	// deleted_at index, which every live row matches.
+	DeliveryID string         `gorm:"type:uuid;not null;index:idx_delivery_results_delivery_id,priority:1" json:"deliveryId"`
+	DeletedAt  gorm.DeletedAt `gorm:"index:idx_delivery_results_delivery_id,priority:2"                    json:"deletedAt,omitzero"`
+
 	AttemptNum   int    `gorm:"not null"             json:"attemptNum"`
 	Success      bool   `json:"success"`
 	StatusCode   int    `json:"statusCode,omitempty"`
@@ -1,9 +1,27 @@
 package database

+import (
+	"time"
+
+	"gorm.io/gorm"
+)
+
 // Event represents a captured webhook event
+//
+//nolint:lll // a struct tag cannot wrap
 type Event struct {
 	BaseModel

+	// CreatedAt and DeletedAt repeat the BaseModel fields only to index
+	// them for retention, which finds events by age. Its lookups carry
+	// GORM's "deleted_at IS NULL" (see DeliveryResult) and compare
+	// created_at with <, so their index has deleted_at first: SQLite
+	// narrows by a < only on the last column it uses. Its final delete
+	// has no deleted_at condition and uses the index on created_at
+	// alone. The other tables keep the unindexed BaseModel created_at.
+	CreatedAt time.Time      `gorm:"index;index:idx_events_deleted_at_created_at,priority:2" json:"createdAt"`
+	DeletedAt gorm.DeletedAt `gorm:"index:idx_events_deleted_at_created_at,priority:1"       json:"deletedAt,omitzero"`
+
 	WebhookID    string `gorm:"type:uuid;not null" json:"webhookId"`
 	EntrypointID string `gorm:"type:uuid;not null" json:"entrypointId"`

@@ -0,0 +1,240 @@
+package database_test
+
+import (
+	"context"
+	"io/fs"
+	"net/http"
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/google/uuid"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"go.uber.org/fx/fxtest"
+	"sneak.berlin/go/webhooker/internal/config"
+	"sneak.berlin/go/webhooker/internal/database"
+	"sneak.berlin/go/webhooker/internal/globals"
+	"sneak.berlin/go/webhooker/internal/logger"
+)
+
+// ownerOnly is the mode every SQLite file the service owns must have.
+// Spelled out rather than referencing database.SQLiteFilePerm so the
+// test fails if the constant itself is loosened.
+const ownerOnly fs.FileMode = 0o600
+
+// requireOwnerOnly asserts that path exists and is readable and
+// writable by its owner and by nobody else.
+func requireOwnerOnly(t *testing.T, path string) {
+	t.Helper()
+
+	info, err := os.Stat(path)
+	require.NoError(t, err, "%s must exist", path)
+	assert.Equal(
+		t,
+		ownerOnly,
+		info.Mode().Perm(),
+		"%s holds credentials and must not be readable by "+
+			"anyone but its owner",
+		path,
+	)
+}
+
+// requireDatabaseSetOwnerOnly asserts the mode of a database file and
+// of both WAL sidecars. The sidecars carry the same rows as the
+// database, so tightening only the main file fixes nothing.
+func requireDatabaseSetOwnerOnly(t *testing.T, dbPath string) {
+	t.Helper()
+
+	requireOwnerOnly(t, dbPath)
+	requireOwnerOnly(t, dbPath+"-wal")
+	requireOwnerOnly(t, dbPath+"-shm")
+}
+
+// TestMainDatabaseFilesAreOwnerOnly covers the tier the defect was
+// reported against: webhooker.db holds targets.config in plaintext —
+// bearer tokens, API keys, Slack webhook URLs — and the session
+// encryption key.
+func TestMainDatabaseFilesAreOwnerOnly(t *testing.T) {
+	t.Parallel()
+
+	lc := fxtest.NewLifecycle(t)
+
+	l, err := logger.New(lc, logger.LoggerParams{
+		Globals: &globals.Globals{
+			Appname: testAppname,
+			Version: testVersion,
+		},
+	})
+	require.NoError(t, err)
+
+	// A directory the application creates itself, not one t.TempDir
+	// made at 0700, so the mode below is the application's.
+	dataDir := filepath.Join(t.TempDir(), "data")
+
+	db, err := database.New(lc, database.DatabaseParams{
+		Config: &config.Config{DataDir: dataDir},
+		Logger: l,
+	})
+	require.NoError(t, err)
+
+	ctx := context.Background()
+	require.NoError(t, lc.Start(ctx))
+
+	defer func() { require.NoError(t, lc.Stop(ctx)) }()
+
+	// Write through the real model so the WAL is populated and both
+	// sidecars are on disk while the handle is open.
+	require.NoError(t, db.DB().Create(&database.Webhook{
+		Name: testWebhookName,
+	}).Error)
+
+	requireDatabaseSetOwnerOnly(
+		t, filepath.Join(dataDir, database.MainDBFileName),
+	)
+
+	// The data directory grants nothing to `other`. Asserted as a
+	// property rather than as an exact 0750, because MkdirAll applies
+	// the ambient umask: the exact mode is the developer's umask as
+	// much as the application's request, and pinning it would make
+	// `make check` pass or fail on where it is run. The group bits are
+	// deliberately left unasserted — deployments may rely on them.
+	info, err := os.Stat(dataDir)
+	require.NoError(t, err)
+	assert.Zero(
+		t,
+		info.Mode().Perm()&0o007,
+		"the data directory must not be world-accessible",
+	)
+}
+
+// TestPerWebhookEventDatabaseFilesAreOwnerOnly covers the events-*.db
+// tier. These carry no credential canaries since
+// https://git.eeqj.de/sneak/webhooker/issues/206, but they hold every
+// received request body and header.
+func TestPerWebhookEventDatabaseFilesAreOwnerOnly(t *testing.T) {
+	t.Parallel()
+
+	mgr, lc := setupTestWebhookDBManager(t)
+	ctx := context.Background()
+	require.NoError(t, lc.Start(ctx))
+
+	defer func() { require.NoError(t, lc.Stop(ctx)) }()
+
+	webhookID := uuid.New().String()
+
+	db, err := mgr.GetDB(webhookID)
+	require.NoError(t, err)
+
+	require.NoError(t, db.Create(&database.Event{
+		WebhookID:    webhookID,
+		EntrypointID: uuid.New().String(),
+		Method:       http.MethodPost,
+		Body:         "{}",
+	}).Error)
+
+	requireDatabaseSetOwnerOnly(t, mgr.DBPath(webhookID))
+}
+
+// TestArchiveDatabaseFilesAreOwnerOnly covers the archive-*.db tier.
+// internal/delivery builds that path and opens it through OpenSQLite,
+// the same single open path exercised here, so the mode is settled for
+// all three tiers in one place.
+func TestArchiveDatabaseFilesAreOwnerOnly(t *testing.T) {
+	t.Parallel()
+
+	ctx := context.Background()
+	path := filepath.Join(
+		t.TempDir(), "archive-"+uuid.New().String()+".db",
+	)
+
+	sqlDB, err := database.OpenSQLite(path, database.SQLiteModeCreate)
+	require.NoError(t, err)
+
+	defer func() { require.NoError(t, sqlDB.Close()) }()
+
+	_, err = sqlDB.ExecContext(ctx, "create table t (id integer)")
+	require.NoError(t, err)
+
+	requireDatabaseSetOwnerOnly(t, path)
+}
+
+// TestOpenSQLiteTightensFilesLeftWorldReadable is the upgrade case: a
+// data directory an earlier build left at 0644, including a
+// developer's own scratch directory, is fixed when it is opened rather
+// than staying exposed until it is recreated.
+func TestOpenSQLiteTightensFilesLeftWorldReadable(t *testing.T) {
+	t.Parallel()
+
+	dir := t.TempDir()
+	path := filepath.Join(dir, database.MainDBFileName)
+
+	// A database and both sidecars as the pre-fix build left them.
+	for _, p := range []string{path, path + "-wal", path + "-shm"} {
+		require.NoError(t, os.WriteFile(p, nil, 0o644)) //nolint:gosec // the mode under test
+	}
+
+	sqlDB, err := database.OpenSQLite(path, database.SQLiteModeCreate)
+	require.NoError(t, err)
+
+	require.NoError(t, sqlDB.Close())
+
+	requireDatabaseSetOwnerOnly(t, path)
+}
+
+// TestOpenSQLiteExistingModeDoesNotCreateTheFile guards the mechanism
+// the fix uses: OpenSQLite now creates the database file itself, and
+// must not do so for a caller that asked for an existing database. An
+// empty file materialized here would turn a missing-database error
+// into a silently empty one.
+func TestOpenSQLiteExistingModeDoesNotCreateTheFile(t *testing.T) {
+	t.Parallel()
+
+	ctx := context.Background()
+	path := filepath.Join(t.TempDir(), "absent.db")
+
+	sqlDB, err := database.OpenSQLite(path, database.SQLiteModeExisting)
+	if err == nil {
+		// sql.Open is lazy: force the connection that fails.
+		require.Error(t, sqlDB.PingContext(ctx))
+		require.NoError(t, sqlDB.Close())
+	}
+
+	_, statErr := os.Stat(path)
+	assert.ErrorIs(t, statErr, fs.ErrNotExist)
+}
+
+// TestReopenAfterRestartKeepsFilesOwnerOnly is the restart case: a
+// process that closed its files must be able to open them again at
+// 0600, including through a gorm handle, and the sidecars must come
+// back at 0600 too rather than at SQLite's own default.
+func TestReopenAfterRestartKeepsFilesOwnerOnly(t *testing.T) {
+	t.Parallel()
+
+	ctx := context.Background()
+	dir := t.TempDir()
+	path := filepath.Join(dir, database.MainDBFileName)
+
+	first, err := database.OpenSQLite(path, database.SQLiteModeCreate)
+	require.NoError(t, err)
+
+	_, err = first.ExecContext(ctx, "create table t (id integer)")
+	require.NoError(t, err)
+	require.NoError(t, first.Close())
+
+	second, err := database.OpenSQLite(path, database.SQLiteModeCreate)
+	require.NoError(t, err)
+
+	defer func() { require.NoError(t, second.Close()) }()
+
+	_, err = second.ExecContext(ctx, "insert into t (id) values (1)")
+	require.NoError(t, err)
+
+	requireDatabaseSetOwnerOnly(t, path)
+
+	var got int
+
+	require.NoError(t,
+		second.QueryRowContext(ctx, "select id from t").Scan(&got))
+	assert.Equal(t, 1, got)
+}
@@ -0,0 +1,252 @@
+package database
+
+import (
+	"database/sql"
+	"errors"
+	"fmt"
+	"io/fs"
+	"net/url"
+	"os"
+	"time"
+
+	_ "modernc.org/sqlite" // Pure Go SQLite driver
+)
+
+// Every SQLite file this service opens — the main database, the
+// per-webhook event databases, and the archive databases — is opened
+// through OpenSQLite, so the durability settings below are properties
+// of the service rather than of one call site.
+//
+// modernc.org/sqlite installs no busy handler and issues no pragmas of
+// its own: it executes only the pragmas named in explicit `_pragma=`
+// DSN parameters, and gorm.io/driver/sqlite adds none when it is
+// handed an existing *sql.DB. Every setting therefore has to be
+// spelled out here or it is simply not in effect.
+// SQLite URI open modes.
+const (
+	// SQLiteModeCreate creates the database file when it is missing.
+	SQLiteModeCreate = "rwc"
+
+	// SQLiteModeExisting requires the file to exist already.
+	SQLiteModeExisting = "rw"
+)
+
+const (
+	// SQLiteBusyTimeout is how long SQLite retries a lock conflict
+	// before returning SQLITE_BUSY.
+	//
+	// Under WAL a reader never blocks a writer, so the only conflict
+	// left is writer against writer: this process's delivery workers
+	// against each other, or against another process holding the write
+	// lock. Those clear in milliseconds. Ten seconds is far above that
+	// and still well inside the receiver's request budget, so an
+	// inbound webhook waits rather than being rejected with a 500.
+	SQLiteBusyTimeout = 10 * time.Second
+
+	// sqliteMaxOpenConns bounds the connection pool for one database
+	// file.
+	//
+	// The pool needs a bound at all because database/sql cannot detect
+	// a connection left mid-transaction: modernc.org/sqlite implements
+	// neither driver.Validator nor driver.SessionResetter, so a
+	// connection whose COMMIT failed is returned to the pool with its
+	// transaction still open and handed out again indefinitely. That is
+	// what turned four `database is locked` errors into 593
+	// `cannot start a transaction within a transaction` in
+	// https://git.eeqj.de/sneak/webhooker/issues/256.
+	//
+	// Four is above the one writer SQLite allows at a time, so reads
+	// still proceed while a write is in flight, and low enough that
+	// contention is resolved by the busy handler rather than by piling
+	// up connections against a lock only one of them can hold.
+	sqliteMaxOpenConns = 4
+
+	// sqliteMaxIdleConns keeps the pool warm without holding every
+	// connection open through an idle period.
+	sqliteMaxIdleConns = 2
+
+	// sqliteConnMaxLifetime and sqliteConnMaxIdleTime retire pooled
+	// connections on a schedule. With _txlock=immediate a failed
+	// COMMIT should no longer be reachable, but these bound the damage
+	// if one happens anyway: a poisoned connection is closed and
+	// replaced within the lifetime instead of wedging the file until
+	// the process restarts.
+	sqliteConnMaxLifetime = 5 * time.Minute
+	sqliteConnMaxIdleTime = time.Minute
+)
+
+// SQLiteFilePerm is the mode every SQLite file this service owns is
+// created with and held at: owner read/write, nothing for group or
+// other.
+//
+// These files hold credentials in plaintext. The main database stores
+// `targets.config` — bearer tokens, API keys, Slack webhook URLs — and
+// the session encryption key. SQLite left to itself creates them 0644
+// (see reserveSQLiteFile), which made the 0750 data directory the only
+// barrier; a bind-mounted directory supplied at 0755 removes it and
+// every local user on the host can read every stored credential.
+//
+// This is a file-mode fix and not encryption at rest. An unattended
+// process needs a key it can read without a human, so the key lands
+// beside the data and an attacker who can read the database can read
+// it too. See https://git.eeqj.de/sneak/webhooker/issues/212.
+const SQLiteFilePerm fs.FileMode = 0o600
+
+// reserveSQLiteFile puts path at SQLiteFilePerm before the driver ever
+// touches it, and tightens any sidecar already on disk.
+//
+// The mode has to be settled here rather than by a chmod after opening,
+// because SQLite picks it: robust_open substitutes
+// SQLITE_DEFAULT_FILE_PERMISSIONS (0644) whenever it is handed mode 0,
+// and findCreateFileMode yields 0 for a main database opened by URI
+// with no `modeof` parameter. A chmod afterwards would leave a window
+// in which the credentials are on disk world-readable.
+//
+// Creating the file ourselves also settles the sidecars, which is the
+// half that could quietly not work. SQLite does not create those at a
+// mode we choose — it derives both from the main database file:
+// `-wal` through findCreateFileMode, which stats the path with the
+// suffix stripped, and `-shm` in unixOpenSharedMemory from an fstat of
+// the already-open database descriptor. A main file at 0600 therefore
+// produces sidecars at 0600. A zero-length file is a valid empty
+// database, so reserving it changes nothing else.
+//
+// create says whether the caller is opening in a mode that may create
+// the database. When it is false a missing file is left missing, so
+// SQLite still reports the absence rather than this function
+// materializing an empty database the caller asked not to create.
+//
+// Chmod of a file that already exists is what tightens a data
+// directory an earlier build left at 0644 — including a developer's
+// own scratch directory — without any migration machinery.
+func reserveSQLiteFile(path string, create bool) error {
+	if create {
+		// gosec G304: the path is the database file the caller asked
+		// to open, and the driver is about to open the same path
+		// anyway. Creating it here is what fixes its mode.
+		f, err := os.OpenFile( //nolint:gosec // see above
+			path, os.O_RDWR|os.O_CREATE, SQLiteFilePerm,
+		)
+		if err != nil {
+			return fmt.Errorf("creating %s: %w", path, err)
+		}
+
+		err = f.Close()
+		if err != nil {
+			return fmt.Errorf("closing %s: %w", path, err)
+		}
+	}
+
+	// O_CREATE leaves an existing file's mode alone, and umask can only
+	// have narrowed a new one. Chmod settles both cases at exactly
+	// SQLiteFilePerm.
+	for _, p := range append(
+		[]string{path}, sqliteSidecarPaths(path)...,
+	) {
+		err := os.Chmod(p, SQLiteFilePerm)
+		if err != nil && !errors.Is(err, fs.ErrNotExist) {
+			return fmt.Errorf("securing %s: %w", p, err)
+		}
+	}
+
+	return nil
+}
+
+// sqliteSidecarPaths returns the files SQLite maintains beside a
+// database under WAL. They carry the same rows as the database itself,
+// so a fix that tightens only the main file has fixed nothing.
+func sqliteSidecarPaths(path string) []string {
+	return []string{path + "-wal", path + "-shm"}
+}
+
+// SQLiteDSN builds the connection string for one database file.
+//
+// mode is the SQLite URI open mode: "rwc" to create the file when it
+// is missing, "rw" to require that it already exists.
+//
+// Three settings carry the fix for
+// https://git.eeqj.de/sneak/webhooker/issues/256 and none of them is
+// optional:
+//
+//   - journal_mode=WAL, so a reader — an operator running
+//     `sqlite3 <db> .dump` over their own data — takes a snapshot
+//     instead of blocking every writer behind it.
+//
+//   - busy_timeout, so a writer that does meet a lock waits for it.
+//     Without one SQLite gives up immediately; nothing above it
+//     retries.
+//
+//   - _txlock=immediate, so every transaction takes the write lock at
+//     BEGIN. A deferred transaction acquires it lazily on its first
+//     write, and that upgrade returns SQLITE_BUSY *without* consulting
+//     the busy handler, because SQLite cannot block a transaction that
+//     may already hold a read snapshot. Such a COMMIT then fails while
+//     the transaction stays open on the connection. A busy timeout
+//     alone does not prevent this; BEGIN IMMEDIATE does, by putting
+//     the wait somewhere the handler applies.
+//
+// Note what is absent: `cache=shared`. Under a shared cache an
+// in-process conflict is reported as SQLITE_LOCKED rather than
+// SQLITE_BUSY, and the busy handler does not retry SQLITE_LOCKED — so
+// leaving it in would have defeated the busy timeout for exactly the
+// contention this service generates. Dropping it is part of the fix,
+// not housekeeping.
+//
+// synchronous is deliberately left at SQLite's default of FULL: this
+// is a webhook receiver whose one promise is that an event it answered
+// 200 for is durable.
+// The order of the _pragma parameters is load-bearing.
+// modernc.org/sqlite executes them in the order they appear, on every
+// new connection, before the connection is handed to the pool. Setting
+// journal_mode first means that pragma itself runs with no busy
+// handler installed: the pool opens connections lazily, so the moment
+// a new one is created is a moment the database is under load, and
+// PRAGMA journal_mode takes a lock. It would fail immediately with
+// SQLITE_BUSY and fail the query that caused the connection to be
+// opened. busy_timeout is therefore set first, so every pragma after
+// it — and the whole life of the connection — is covered.
+func SQLiteDSN(path, mode string) string {
+	q := url.Values{}
+	q.Set("mode", mode)
+	q.Set("_txlock", "immediate")
+	q.Add(
+		"_pragma",
+		fmt.Sprintf(
+			"busy_timeout(%d)",
+			SQLiteBusyTimeout.Milliseconds(),
+		),
+	)
+	q.Add("_pragma", "journal_mode(WAL)")
+
+	return "file:" + path + "?" + q.Encode()
+}
+
+// OpenSQLite opens the SQLite file at path with the service's
+// durability settings and pool bounds applied. mode is the SQLite URI
+// open mode ("rwc" or "rw").
+//
+// The file and its WAL sidecars are settled at SQLiteFilePerm before
+// the driver sees the path; see reserveSQLiteFile.
+//
+// The handle is returned rather than a *gorm.DB because the callers
+// wrap it in gorm themselves with their own logger.
+func OpenSQLite(path, mode string) (*sql.DB, error) {
+	err := reserveSQLiteFile(path, mode == SQLiteModeCreate)
+	if err != nil {
+		return nil, err
+	}
+
+	sqlDB, err := sql.Open("sqlite", SQLiteDSN(path, mode))
+	if err != nil {
+		return nil, fmt.Errorf(
+			"opening sqlite database %s: %w", path, err,
+		)
+	}
+
+	sqlDB.SetMaxOpenConns(sqliteMaxOpenConns)
+	sqlDB.SetMaxIdleConns(sqliteMaxIdleConns)
+	sqlDB.SetConnMaxLifetime(sqliteConnMaxLifetime)
+	sqlDB.SetConnMaxIdleTime(sqliteConnMaxIdleTime)
+
+	return sqlDB, nil
+}
@@ -0,0 +1,178 @@
+package database_test
+
+import (
+	"context"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/google/uuid"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"gorm.io/gorm"
+	"sneak.berlin/go/webhooker/internal/database"
+)
+
+// livePragma reads a pragma off a live handle. Reading the DSN back
+// would prove only that the string was built; these tests assert that
+// SQLite actually applied it.
+func livePragma(t *testing.T, db *gorm.DB, name string) string {
+	t.Helper()
+
+	var v string
+
+	row := db.Raw("pragma " + name).Row()
+	require.NoError(t, row.Scan(&v))
+
+	return v
+}
+
+func TestSQLiteDSNCarriesTheDurabilitySettings(t *testing.T) {
+	t.Parallel()
+
+	dsn := database.SQLiteDSN(
+		"/var/lib/webhooker/webhooker.db",
+		database.SQLiteModeCreate,
+	)
+
+	assert.Contains(t, dsn, "journal_mode%28WAL%29")
+	assert.Contains(t, dsn, "busy_timeout%2810000%29")
+	assert.Contains(t, dsn, "_txlock=immediate")
+	assert.Contains(t, dsn, "mode=rwc")
+
+	// busy_timeout must come first. The driver runs these in order on
+	// every new connection, and PRAGMA journal_mode takes a lock — a
+	// connection opened while the database is busy would fail on that
+	// pragma, with no busy handler yet installed to wait it out.
+	assert.Less(
+		t,
+		strings.Index(dsn, "busy_timeout"),
+		strings.Index(dsn, "journal_mode"),
+		"busy_timeout must be applied before journal_mode",
+	)
+
+	// cache=shared turns an in-process conflict into SQLITE_LOCKED,
+	// which the busy handler does not retry. It must never come back.
+	// See https://git.eeqj.de/sneak/webhooker/issues/256.
+	assert.NotContains(t, strings.ToLower(dsn), "cache=shared")
+}
+
+// TestPerWebhookDBAppliesPragmasOnALiveHandle is the check the issue
+// asks for by name: the settings are confirmed by querying the running
+// database, not by inspecting the connection string.
+func TestPerWebhookDBAppliesPragmasOnALiveHandle(t *testing.T) {
+	t.Parallel()
+
+	mgr, lc := setupTestWebhookDBManager(t)
+	ctx := context.Background()
+	require.NoError(t, lc.Start(ctx))
+
+	defer func() { require.NoError(t, lc.Stop(ctx)) }()
+
+	webhookID := uuid.New().String()
+
+	db, err := mgr.GetDB(webhookID)
+	require.NoError(t, err)
+
+	assert.Equal(
+		t, "wal",
+		strings.ToLower(livePragma(t, db, "journal_mode")),
+	)
+	assert.Equal(
+		t, "10000", livePragma(t, db, "busy_timeout"),
+	)
+}
+
+func TestMainDBAppliesPragmasOnALiveHandle(t *testing.T) {
+	t.Parallel()
+
+	ctx := context.Background()
+	dir := t.TempDir()
+
+	sqlDB, err := database.OpenSQLite(
+		filepath.Join(dir, database.MainDBFileName),
+		database.SQLiteModeCreate,
+	)
+	require.NoError(t, err)
+
+	defer func() { require.NoError(t, sqlDB.Close()) }()
+
+	var journal string
+
+	require.NoError(t, sqlDB.
+		QueryRowContext(ctx, "pragma journal_mode").
+		Scan(&journal))
+	assert.Equal(t, "wal", strings.ToLower(journal))
+
+	var busy string
+
+	require.NoError(t, sqlDB.
+		QueryRowContext(ctx, "pragma busy_timeout").
+		Scan(&busy))
+	assert.Equal(t, "10000", busy)
+}
+
+// TestConcurrentReaderDoesNotBlockWrites is the unit-scale form of the
+// reproduction in
+// https://git.eeqj.de/sneak/webhooker/issues/256: an operator's
+// long-held read of their own data used to make every concurrent write
+// fail. Under WAL the reader takes a snapshot and the writes proceed.
+func TestConcurrentReaderDoesNotBlockWrites(t *testing.T) {
+	t.Parallel()
+
+	mgr, lc := setupTestWebhookDBManager(t)
+	ctx := context.Background()
+	require.NoError(t, lc.Start(ctx))
+
+	defer func() { require.NoError(t, lc.Stop(ctx)) }()
+
+	webhookID := uuid.New().String()
+
+	db, err := mgr.GetDB(webhookID)
+	require.NoError(t, err)
+
+	// A second handle on the same file, holding a read transaction
+	// open across every write below — what `sqlite3 <db> .dump` is.
+	readerSQL, err := database.OpenSQLite(
+		mgr.DBPath(webhookID), database.SQLiteModeExisting,
+	)
+	require.NoError(t, err)
+
+	defer func() { require.NoError(t, readerSQL.Close()) }()
+
+	readerConn, err := readerSQL.Conn(ctx)
+	require.NoError(t, err)
+
+	defer func() { require.NoError(t, readerConn.Close()) }()
+
+	_, err = readerConn.ExecContext(ctx, "begin deferred")
+	require.NoError(t, err)
+
+	_, err = readerConn.ExecContext(
+		ctx, "select count(*) from events",
+	)
+	require.NoError(t, err)
+
+	for range 25 {
+		err = db.Transaction(func(tx *gorm.DB) error {
+			return tx.Create(&database.Event{
+				WebhookID:    webhookID,
+				EntrypointID: uuid.New().String(),
+				Method:       "POST",
+				Body:         "{}",
+			}).Error
+		})
+		require.NoError(t, err)
+	}
+
+	_, err = readerConn.ExecContext(ctx, "commit")
+	require.NoError(t, err)
+
+	var count int64
+
+	require.NoError(
+		t,
+		db.Model(&database.Event{}).Count(&count).Error,
+	)
+	assert.Equal(t, int64(25), count)
+}
@@ -2,7 +2,6 @@ package database

 import (
 	"context"
-	"database/sql"
 	"errors"
 	"fmt"
 	"log/slog"
@@ -14,6 +13,7 @@ import (
 	"gorm.io/driver/sqlite"
 	"gorm.io/gorm"
 	"sneak.berlin/go/webhooker/internal/config"
+	"sneak.berlin/go/webhooker/internal/datadir"
 	"sneak.berlin/go/webhooker/internal/gormlog"
 	"sneak.berlin/go/webhooker/internal/logger"
 )
@@ -41,6 +41,11 @@ type WebhookDBManager struct {
 	dataDir string
 	dbs     sync.Map // map[webhookID]*gorm.DB
 	log     *slog.Logger
+
+	// mu is held while a database is opened, deleted, or closed, so
+	// each file has at most one open handle. Reading an already cached
+	// handle does not take it.
+	mu sync.Mutex
 }

 // NewWebhookDBManager creates a new WebhookDBManager and
@@ -54,8 +59,9 @@ func NewWebhookDBManager(
 		log:     params.Logger.Get(),
 	}

-	// Create data directory if it doesn't exist
-	err := os.MkdirAll(m.dataDir, dataDirPerm)
+	// Create data directory if it doesn't exist. datadir.DirPerm is the
+	// single source of the directory mode; either package may run first.
+	err := os.MkdirAll(m.dataDir, datadir.DirPerm)
 	if err != nil {
 		return nil, fmt.Errorf(
 			"creating data directory %s: %w",
@@ -85,43 +91,39 @@ func (m *WebhookDBManager) GetDB(
 ) (*gorm.DB, error) {
 	// Fast path: already open
 	if val, ok := m.dbs.Load(webhookID); ok {
-		cachedDB, castOK := val.(*gorm.DB)
-		if !castOK {
-			return nil, fmt.Errorf(
-				"%w for webhook %s",
-				errInvalidCachedDBType,
-				webhookID,
-			)
-		}
+		return asGormDB(val, webhookID)
+	}

-		return cachedDB, nil
+	// Slow path: open the database under the lock, looking in the
+	// cache again first. A caller that raced another one here then
+	// waits for its handle instead of opening a second one.
+	m.mu.Lock()
+	defer m.mu.Unlock()
+
+	if val, ok := m.dbs.Load(webhookID); ok {
+		return asGormDB(val, webhookID)
 	}

-	// Slow path: open/create the database
 	db, err := m.openDB(webhookID)
 	if err != nil {
 		return nil, err
 	}

-	// Store it; if another goroutine beat us, close ours
-	actual, loaded := m.dbs.LoadOrStore(webhookID, db)
-	if loaded {
-		// Another goroutine created it first; close our duplicate
-		sqlDB, closeErr := db.DB()
-		if closeErr == nil {
-			_ = sqlDB.Close()
-		}
+	m.dbs.Store(webhookID, db)

-		existingDB, castOK := actual.(*gorm.DB)
-		if !castOK {
-			return nil, fmt.Errorf(
-				"%w for webhook %s",
-				errInvalidCachedDBType,
-				webhookID,
-			)
-		}
+	return db, nil
+}

-		return existingDB, nil
+// asGormDB returns a value read from the cache as the database
+// handle it is.
+func asGormDB(val any, webhookID string) (*gorm.DB, error) {
+	db, ok := val.(*gorm.DB)
+	if !ok {
+		return nil, fmt.Errorf(
+			"%w for webhook %s",
+			errInvalidCachedDBType,
+			webhookID,
+		)
 	}

 	return db, nil
@@ -152,6 +154,11 @@ func (m *WebhookDBManager) DBExists(
 func (m *WebhookDBManager) DeleteDB(
 	webhookID string,
 ) error {
+	// Held until the files are gone, so GetDB cannot open the file
+	// again between the close and the removal.
+	m.mu.Lock()
+	defer m.mu.Unlock()
+
 	// Close and remove from cache
 	if val, ok := m.dbs.LoadAndDelete(webhookID); ok {
 		if gormDB, castOK := val.(*gorm.DB); castOK {
@@ -185,6 +192,11 @@ func (m *WebhookDBManager) DeleteDB(
 // CloseAll closes all open per-webhook database connections.
 // Called during application shutdown.
 func (m *WebhookDBManager) CloseAll() error {
+	// An open already under way finishes and is cached first, so it
+	// is closed here rather than cached after this loop has passed.
+	m.mu.Lock()
+	defer m.mu.Unlock()
+
 	var lastErr error

 	m.dbs.Range(func(key, value any) bool {
@@ -234,12 +246,11 @@ func (m *WebhookDBManager) openDB(
 	webhookID string,
 ) (*gorm.DB, error) {
 	path := m.dbPath(webhookID)
-	dbURL := fmt.Sprintf(
-		"file:%s?cache=shared&mode=rwc",
-		path,
-	)

-	sqlDB, err := sql.Open("sqlite", dbURL)
+	// See sqlite_open.go: WAL, a busy timeout, immediate-transaction
+	// locking, and a bounded pool, all of which this file needs most —
+	// it is the one every delivery worker writes to concurrently.
+	sqlDB, err := OpenSQLite(path, SQLiteModeCreate)
 	if err != nil {
 		return nil, fmt.Errorf(
 			"opening webhook database %s: %w",
@@ -1,10 +1,14 @@
 package database_test

 import (
+	"bytes"
 	"context"
+	"log/slog"
 	"net/http"
 	"os"
 	"path/filepath"
+	"strings"
+	"sync"
 	"testing"

 	"github.com/google/uuid"
@@ -104,6 +108,54 @@ func TestWebhookDBManager_CreateAndGetDB(t *testing.T) {
 	assert.Equal(t, `{"test": true}`, readEvent.Body)
 }

+// Many callers ask for one webhook's database at the same moment,
+// before it is cached. Only one of them may open the file; the others
+// must wait for its handle. openDB logs one "opened per-webhook
+// database" line per open, and those lines are what is counted.
+func TestWebhookDBManager_ConcurrentFirstTouchOpensOnce(t *testing.T) {
+	t.Parallel()
+
+	var logs bytes.Buffer
+
+	mgr := database.NewTestWebhookDBManagerWithLogger(
+		t.TempDir(),
+		slog.New(slog.NewTextHandler(&logs, nil)),
+	)
+
+	t.Cleanup(func() { assert.NoError(t, mgr.CloseAll()) })
+
+	webhookID := uuid.New().String()
+
+	const callers = 16
+
+	start := make(chan struct{})
+	handles := make([]*gorm.DB, callers)
+	errs := make([]error, callers)
+
+	var wg sync.WaitGroup
+
+	for i := range callers {
+		wg.Go(func() {
+			<-start
+
+			handles[i], errs[i] = mgr.GetDB(webhookID)
+		})
+	}
+
+	close(start)
+	wg.Wait()
+
+	for i := range callers {
+		require.NoError(t, errs[i])
+		assert.Same(t, handles[0], handles[i])
+	}
+
+	assert.Equal(
+		t, 1,
+		strings.Count(logs.String(), "opened per-webhook database"),
+	)
+}
+
 func TestWebhookDBManager_DeleteDB(t *testing.T) {
 	t.Parallel()

@@ -29,9 +29,11 @@ import (
 // process that was killed with SIGKILL blocks nothing.
 const LockFileName = "webhooker.lock"

-// dirPerm is the mode Acquire creates DATA_DIR with. It matches what
-// internal/database uses, since whichever runs first creates it.
-const dirPerm = 0o750
+// DirPerm is the mode DATA_DIR is created with. It is the single
+// source of that mode: internal/database consumes it rather than
+// keeping its own copy, so the two packages that both create the
+// directory cannot drift into disagreeing about its permissions.
+const DirPerm = 0o750

 // ErrLocked reports that another live process holds the data
 // directory. Callers that need to know whether a deployment is running
@@ -64,7 +66,7 @@ func Acquire(dir string) (*Lock, error) {
 		return nil, ErrNoDir
 	}

-	err := os.MkdirAll(dir, dirPerm)
+	err := os.MkdirAll(dir, DirPerm)
 	if err != nil {
 		return nil, fmt.Errorf(
 			"creating data directory %s: %w", dir, err,
@@ -76,12 +76,20 @@ func (cb *CircuitBreaker) Allow() bool {
 	}
 }

-// CooldownRemaining returns how much time is left before
-// an open circuit transitions to half-open.
+// CooldownRemaining returns how long a delivery that Allow refused
+// should wait before it is tried again. Closed, it returns zero.
+// Open, it returns what is left of the cooldown, or zero once that
+// has passed. Half-open, it returns the whole cooldown: the one
+// probe delivery is still in flight, and if it fails the circuit
+// reopens for that long.
 func (cb *CircuitBreaker) CooldownRemaining() time.Duration {
 	cb.mu.Lock()
 	defer cb.mu.Unlock()

+	if cb.state == CircuitHalfOpen {
+		return cb.cooldown
+	}
+
 	if cb.state != CircuitOpen {
 		return 0
 	}
@@ -267,7 +267,7 @@ func TestCircuitBreaker_CooldownRemaining_ClosedReturnsZero(
 	)
 }

-func TestCircuitBreaker_CooldownRemaining_HalfOpenReturnsZero(
+func TestCircuitBreaker_CooldownRemaining_HalfOpenReturnsCooldown(
 	t *testing.T,
 ) {
 	t.Parallel()
@@ -282,9 +282,11 @@ func TestCircuitBreaker_CooldownRemaining_HalfOpenReturnsZero(

 	require.True(t, cb.Allow())

-	assert.Equal(t, time.Duration(0),
+	// The cooldown newShortCooldownCB gives the breaker.
+	assert.Equal(t, 50*time.Millisecond,
 		cb.CooldownRemaining(),
-		"half-open circuit should have zero cooldown remaining",
+		"a delivery refused while half-open should wait "+
+			"a whole cooldown",
 	)
 }

@@ -2,7 +2,6 @@ package delivery_test

 import (
 	"context"
-	"database/sql"
 	"encoding/json"
 	"fmt"
 	"io"
@@ -70,11 +69,12 @@ func iMainDB(t *testing.T) *gorm.DB {
 		t.TempDir(), "main-test.db",
 	)

-	dsn := fmt.Sprintf(
-		"file:%s?cache=shared&mode=rwc", dbPath,
+	// Opened the way the service opens the main database, so these
+	// tests cannot pass against journal and locking settings
+	// production does not use.
+	sqlDB, err := database.OpenSQLite(
+		dbPath, database.SQLiteModeCreate,
 	)
-
-	sqlDB, err := sql.Open("sqlite", dsn)
 	require.NoError(t, err)

 	t.Cleanup(func() { _ = sqlDB.Close() })
@@ -377,6 +377,17 @@ func TestProcessRetryTask_SuccessfulRetry(t *testing.T) {

 	bodyStr := event.Body
 	cfg := iHTTPConfig(ts.URL)
+
+	// The target row exists because the engine confirms a scheduled
+	// retry's target has not been deleted before it runs it. A retry
+	// task whose target id names no row at all is a state the service
+	// does not produce: the handler read that target to build the
+	// task. See https://git.eeqj.de/sneak/webhooker/issues/107.
+	iCreateTarget(
+		t, s.MainDB, targetID, s.WebhookID, "retry-target",
+		database.TargetTypeHTTP, cfg, 5,
+	)
+
 	task := iTask(
 		d, event, s.WebhookID, targetID,
 		"retry-target", cfg, 5, 2, &bodyStr,
@@ -456,6 +467,12 @@ func TestProcessRetryTask_LargeBody_FetchFromDB(
 	)

 	cfg := iHTTPConfig(ts.URL)
+
+	iCreateTarget(
+		t, s.MainDB, targetID, s.WebhookID, "retry-large",
+		database.TargetTypeHTTP, cfg, 5,
+	)
+
 	task := iTask(
 		d, event, s.WebhookID, targetID,
 		"retry-large", cfg, 5, 2, nil,
@@ -558,6 +575,12 @@ func TestWorkerLifecycle_ProcessesRetryChannel(

 	bodyStr := event.Body
 	cfg := iHTTPConfig(ts.URL)
+
+	iCreateTarget(
+		t, s.MainDB, targetID, s.WebhookID, "retry-chan-test",
+		database.TargetTypeHTTP, cfg, 5,
+	)
+
 	task := iTask(
 		d, event, s.WebhookID, targetID,
 		"retry-chan-test", cfg, 5, 2, &bodyStr,
@@ -2,6 +2,8 @@ package delivery_test

 import (
 	"context"
+	"fmt"
+	"path/filepath"
 	"testing"
 	"time"

@@ -269,3 +271,88 @@ func TestEngine_StopHookHonoursStopTimeout(t *testing.T) {

 	requireStopHookExpires(t, lc.hooks[0], "delivery engine")
 }
+
+// deliverToArchive runs one delivery to a database target through
+// the running engine and returns the webhook's archive file path.
+// The archive writer holds the file open afterwards.
+func deliverToArchive(t *testing.T, s iSetup) string {
+	t.Helper()
+
+	deliveryID, task := seedLogTask(t, s)
+	task.TargetType = database.TargetTypeDatabase
+
+	s.Engine.Notify([]delivery.Task{task})
+
+	iWaitForDelivered(t, s.WebhookDB, deliveryID)
+
+	return filepath.Join(
+		filepath.Dir(s.DBMgr.DBPath(s.WebhookID)),
+		fmt.Sprintf("archive-%s.db", s.WebhookID),
+	)
+}
+
+// TestEngine_StopHookClosesArchives is the regression test for an
+// archive split across two files by a clean stop. The engine never
+// closed its archive writers, so after a stop the archived rows
+// could sit in archive-{id}.db-wal while archive-{id}.db held no
+// table at all, and copying the .db on its own gave an empty
+// database.
+func TestEngine_StopHookClosesArchives(t *testing.T) {
+	t.Parallel()
+
+	s := newISetup(t)
+
+	lc := startEngineViaHook(t, s.Engine)
+
+	path := deliverToArchive(t, s)
+	require.FileExists(
+		t, path+"-wal",
+		"an open archive should have a -wal for the stop to remove",
+	)
+
+	require.NoError(t, lc.hooks[0].OnStop(context.Background()))
+
+	wals, err := filepath.Glob(
+		filepath.Join(filepath.Dir(path), "archive-*.db-wal"),
+	)
+	require.NoError(t, err)
+	require.Empty(
+		t, wals, "a clean stop must leave no archive -wal behind",
+	)
+
+	// With no -wal beside it, the row can only be in the .db.
+	count, err := countArchivedRows(path)
+	require.NoError(t, err)
+	require.Equal(t, int64(1), count)
+}
+
+// TestEngine_StopHookTimeoutLeavesArchivesOpen covers a stop whose
+// budget runs out while a worker is still running. The archive
+// writers are left open, as a kill would leave them: closing them
+// would wait for any write in progress, and that worker would then
+// open new writers that nothing closes.
+func TestEngine_StopHookTimeoutLeavesArchivesOpen(t *testing.T) {
+	t.Parallel()
+
+	s := newISetup(t)
+
+	lc := startEngineViaHook(t, s.Engine)
+
+	deliverToArchive(t, s)
+
+	release := make(chan struct{})
+
+	t.Cleanup(func() {
+		close(release)
+		s.Engine.EvictWebhook(s.WebhookID)
+	})
+
+	s.Engine.ExportWedgeWorker(release)
+
+	requireStopHookExpires(t, lc.hooks[0], "delivery engine")
+
+	require.True(
+		t, s.Engine.ExportArchiveHandleOpen(s.WebhookID),
+		"a stop that timed out must not close archive writers",
+	)
+}
@@ -3,7 +3,6 @@ package delivery_test
 import (
 	"bytes"
 	"context"
-	"database/sql"
 	"encoding/json"
 	"fmt"
 	"log/slog"
@@ -18,6 +17,7 @@ import (
 	"time"

 	"github.com/google/uuid"
+	"github.com/prometheus/client_golang/prometheus"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 	"gorm.io/driver/sqlite"
@@ -25,6 +25,7 @@ import (
 	_ "modernc.org/sqlite"
 	"sneak.berlin/go/webhooker/internal/database"
 	"sneak.berlin/go/webhooker/internal/delivery"
+	"sneak.berlin/go/webhooker/internal/metrics"
 )

 // testContentType is the event content type used in tests.
@@ -37,11 +38,12 @@ func testWebhookDB(t *testing.T) *gorm.DB {
 		t.TempDir(), "events-test.db",
 	)

-	dsn := fmt.Sprintf(
-		"file:%s?cache=shared&mode=rwc", dbPath,
+	// Opened the way the service opens a per-webhook database, so
+	// these tests cannot pass against journal and locking settings
+	// production does not use.
+	sqlDB, err := database.OpenSQLite(
+		dbPath, database.SQLiteModeCreate,
 	)
-
-	sqlDB, err := sql.Open("sqlite", dsn)
 	require.NoError(t, err)

 	t.Cleanup(func() { _ = sqlDB.Close() })
@@ -894,6 +896,100 @@ func TestDeliverHTTP_CircuitBreakerBlocks(t *testing.T) {
 	)
 }

+// recordingScheduler keeps the delay of every retry it is asked to
+// schedule, and schedules nothing.
+type recordingScheduler struct {
+	delays []time.Duration
+}
+
+func (s *recordingScheduler) ScheduleRetry(
+	_ delivery.Task, delay time.Duration,
+) {
+	s.delays = append(s.delays, delay)
+}
+
+// TestDeliverHTTP_HalfOpenBreakerDelaysQueuedTasks proves that while a
+// half-open breaker's one probe delivery is in flight, every other task
+// for the target is put back with a whole cooldown as its delay rather
+// than none, and that its status is written the first time the breaker
+// turns it away and not on each pass after that.
+func TestDeliverHTTP_HalfOpenBreakerDelaysQueuedTasks(t *testing.T) {
+	t.Parallel()
+
+	db := testWebhookDB(t)
+	e := testEngine(t, 1)
+
+	// Every write of retrying moves the retry counter, so on a registry
+	// this test owns the counter is the number of those writes.
+	reg := prometheus.NewRegistry()
+	e.ExportSetMetrics(metrics.New(reg))
+
+	targetID := uuid.New().String()
+	cb := newShortCooldownCB(t)
+	e.ExportSetCircuitBreaker(targetID, cb)
+
+	for range delivery.ExportDefaultFailureThreshold {
+		cb.RecordFailure()
+	}
+
+	time.Sleep(60 * time.Millisecond)
+
+	require.True(t, cb.Allow(), "the probe delivery should go through")
+	require.Equal(t, delivery.CircuitHalfOpen, cb.State())
+
+	cfg := newHTTPTargetConfig(
+		"http://will-not-be-called.invalid",
+	)
+	sched := &recordingScheduler{}
+
+	const queued, passes = 3, 4
+
+	for range queued {
+		event := seedEvent(t, db, `{"cb":"half-open"}`)
+		dlv := seedDelivery(
+			t, db, event.ID, targetID,
+			database.DeliveryStatusPending,
+		)
+
+		for range passes {
+			// Each pass starts from the stored row, as a retry does.
+			var row database.Delivery
+
+			require.NoError(t, db.First(
+				&row, "id = ?", dlv.ID,
+			).Error)
+
+			fix := buildHTTPFixture(
+				row, event, targetID,
+				"test-cb-half-open", cfg, 5, 1,
+			)
+
+			e.ExportDeliverHTTPWithScheduler(
+				context.TODO(), db, fix.Delivery, fix.Task, sched,
+			)
+		}
+
+		assertDeliveryStatus(t, db, dlv.ID,
+			database.DeliveryStatusRetrying,
+		)
+	}
+
+	require.Len(t, sched.delays, queued*passes)
+
+	for _, delay := range sched.delays {
+		// The cooldown newShortCooldownCB gives the breaker.
+		assert.Equal(t, 50*time.Millisecond, delay,
+			"a task turned away while half-open should wait "+
+				"a whole cooldown",
+		)
+	}
+
+	assert.InDelta(t, float64(queued),
+		mCounter(t, reg, mRetries, mTypeHTTP), 0,
+		"status should be written once per task, not once per pass",
+	)
+}
+
 func TestGetCircuitBreaker_CreatesOnDemand(t *testing.T) {
 	t.Parallel()

@@ -1070,6 +1166,10 @@ func TestIsForwardableHeader(t *testing.T) {
 	assert.False(t,
 		delivery.ExportIsForwardableHeader("Content-Length"),
 	)
+
+	assert.False(t,
+		delivery.ExportIsForwardableHeader("Content-Type"),
+	)
 }

 func TestTruncate(t *testing.T) {
@@ -1151,6 +1251,81 @@ func TestDoHTTPRequest_ForwardsHeaders(t *testing.T) {
 	)
 }

+// The event's stored inbound headers carry the same Content-Type the
+// receiver saved as the event's ContentType, so a delivery could send
+// it twice. It must go out exactly once, with a Content-Type configured
+// on the target winning, then the event's ContentType.
+func TestApplyRequestHeaders_SendsOneContentType(t *testing.T) {
+	t.Parallel()
+
+	cases := map[string]struct {
+		inbound    string
+		event      string
+		configured string
+		want       []string
+	}{
+		"inbound and event agree": {
+			inbound: testContentType,
+			event:   testContentType,
+			want:    []string{testContentType},
+		},
+		"inbound and event disagree": {
+			inbound: "text/plain",
+			event:   testContentType,
+			want:    []string{testContentType},
+		},
+		"event has none": {
+			inbound: testContentType,
+			want:    nil,
+		},
+		"target configures its own": {
+			inbound:    testContentType,
+			event:      testContentType,
+			configured: "application/xml",
+			want:       []string{"application/xml"},
+		},
+	}
+
+	for name, tc := range cases {
+		t.Run(name, func(t *testing.T) {
+			t.Parallel()
+
+			inbound, err := json.Marshal(map[string][]string{
+				headerContentType: {tc.inbound},
+			})
+			require.NoError(t, err)
+
+			cfg := &delivery.HTTPTargetConfig{}
+			if tc.configured != "" {
+				cfg.Headers = map[string]string{
+					headerContentType: tc.configured,
+				}
+			}
+
+			req, err := http.NewRequestWithContext(
+				context.Background(),
+				http.MethodPost,
+				"https://target.example.com/hook",
+				http.NoBody,
+			)
+			require.NoError(t, err)
+
+			delivery.ExportApplyRequestHeaders(
+				req,
+				&database.Event{
+					Headers:     string(inbound),
+					ContentType: tc.event,
+				},
+				cfg,
+			)
+
+			assert.Equal(t,
+				tc.want, req.Header.Values(headerContentType),
+			)
+		})
+	}
+}
+
 func TestProcessDelivery_RoutesToCorrectHandler(
 	t *testing.T,
 ) {
@@ -96,7 +96,15 @@ func TestEventDBHoldsNoTargetRows(t *testing.T) {
 	)
 	assertNoTargetRows(t, dbPath)

-	// A retry.
+	// A retry. Its target exists in the main database, because the
+	// engine confirms a scheduled retry's target has not been
+	// deleted before running it; see
+	// https://git.eeqj.de/sneak/webhooker/issues/107.
+	iCreateTarget(
+		t, s.MainDB, targetID, s.WebhookID, "leaky-target",
+		database.TargetTypeHTTP, cfg, 5,
+	)
+
 	rd := iSeedDelivery(
 		t, s.WebhookDB, event.ID, targetID,
 		database.DeliveryStatusRetrying,
@@ -0,0 +1,442 @@
+package delivery_test
+
+import (
+	"context"
+	"encoding/json"
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"testing"
+	"time"
+
+	"github.com/google/uuid"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"gorm.io/gorm"
+	"sneak.berlin/go/webhooker/internal/database"
+	"sneak.berlin/go/webhooker/internal/delivery"
+)
+
+// tsEventCreatedAt is the receipt time seeded on the events these
+// tests deliver. It is far enough from both the zero time and from
+// now that neither can be mistaken for it.
+func tsEventCreatedAt() time.Time {
+	return time.Date(
+		2026, time.March, 4, 5, 6, 7, 0, time.UTC,
+	)
+}
+
+// tsZeroStamp is what a Slack message renders when the event handed
+// to FormatSlackMessage carries no CreatedAt.
+const tsZeroStamp = "*Timestamp:* `0001-01-01T00:00:00Z`"
+
+// tsEventBody is the body seeded on every event in this file. It is
+// small enough that a Task can inline it.
+const tsEventBody = `{"hello":"world"}`
+
+// tsUndeliverableHook stands in for a Slack incoming webhook on the
+// tests that never send: the config parser requires a URL, but no
+// request is made.
+const tsUndeliverableHook = "https://hooks.slack.com/services/T/B/x"
+
+// tsSink is a stand-in Slack incoming webhook that records the raw
+// body posted to it.
+type tsSink struct {
+	*httptest.Server
+
+	bodies chan []byte
+}
+
+func newTSSink(t *testing.T) *tsSink {
+	t.Helper()
+
+	s := &tsSink{bodies: make(chan []byte, 8)}
+
+	s.Server = httptest.NewServer(http.HandlerFunc(
+		func(w http.ResponseWriter, r *http.Request) {
+			body, _ := io.ReadAll(r.Body)
+
+			select {
+			case s.bodies <- body:
+			default:
+			}
+
+			w.WriteHeader(http.StatusOK)
+		},
+	))
+
+	t.Cleanup(s.Close)
+
+	return s
+}
+
+// text returns the Slack message text from the single payload the
+// sink received.
+func (s *tsSink) text(t *testing.T) string {
+	t.Helper()
+
+	select {
+	case raw := <-s.bodies:
+		t.Logf("raw slack payload: %s", raw)
+
+		var payload struct {
+			Text string `json:"text"`
+		}
+
+		require.NoError(t, json.Unmarshal(raw, &payload))
+
+		return payload.Text
+	case <-time.After(5 * time.Second):
+		t.Fatal("slack sink received no payload")
+
+		return ""
+	}
+}
+
+func tsSlackConfig(t *testing.T, url string) string {
+	t.Helper()
+
+	data, err := json.Marshal(
+		delivery.SlackTargetConfig{WebhookURL: url},
+	)
+	require.NoError(t, err)
+
+	return string(data)
+}
+
+// tsSeedEvent writes an event whose CreatedAt is tsEventCreatedAt
+// rather than the write time, so an assertion on the rendered
+// timestamp cannot pass by accident against "roughly now".
+func tsSeedEvent(
+	t *testing.T, db *gorm.DB, webhookID string,
+) database.Event {
+	t.Helper()
+
+	event := database.Event{
+		WebhookID:    webhookID,
+		EntrypointID: uuid.New().String(),
+		Method:       http.MethodPost,
+		Headers:      `{}`,
+		Body:         tsEventBody,
+		ContentType:  "application/json",
+	}
+	event.ID = uuid.New().String()
+	event.CreatedAt = tsEventCreatedAt()
+	event.UpdatedAt = tsEventCreatedAt()
+
+	require.NoError(t, db.Create(&event).Error)
+
+	var stored database.Event
+
+	require.NoError(t,
+		db.First(&stored, "id = ?", event.ID).Error,
+	)
+	require.Equal(t,
+		tsEventCreatedAt().UTC(), stored.CreatedAt.UTC(),
+		"seeded created_at did not round-trip",
+	)
+
+	return event
+}
+
+// tsSeedTarget writes the slack target row into the main database.
+// The retry path confirms the target still exists before sending.
+func tsSeedTarget(
+	t *testing.T, mainDB *gorm.DB, webhookID, config string,
+) database.Target {
+	t.Helper()
+
+	target := database.Target{
+		WebhookID: webhookID,
+		Name:      "slack-sink",
+		Type:      database.TargetTypeSlack,
+		Config:    config,
+		Active:    true,
+	}
+
+	require.NoError(t, mainDB.Create(&target).Error)
+
+	return target
+}
+
+func tsTask(
+	d database.Delivery,
+	event database.Event,
+	webhookID string,
+	target database.Target,
+	attemptNum int,
+	body *string,
+) delivery.Task {
+	return delivery.Task{
+		DeliveryID:   d.ID,
+		EventID:      event.ID,
+		WebhookID:    webhookID,
+		EntrypointID: event.EntrypointID,
+		TargetID:     target.ID,
+		TargetName:   target.Name,
+		TargetType:   database.TargetTypeSlack,
+		TargetConfig: target.Config,
+		MaxRetries:   0,
+		Method:       event.Method,
+		Headers:      event.Headers,
+		ContentType:  event.ContentType,
+		Body:         body,
+		AttemptNum:   attemptNum,
+	}
+}
+
+func tsAssertRealTimestamp(t *testing.T, text string) {
+	t.Helper()
+
+	assert.NotContains(t, text, tsZeroStamp,
+		"slack message carries the zero timestamp",
+	)
+	assert.Contains(t, text,
+		"*Timestamp:* `"+
+			tsEventCreatedAt().UTC().Format(time.RFC3339)+"`",
+		"slack message does not carry the event's receipt time",
+	)
+}
+
+// tsCase is one end-to-end delivery of a seeded event to a slack
+// sink, over whichever engine path `process` names.
+type tsCase struct {
+	// status is the delivery row's status before the engine runs.
+	// The retry path refuses a delivery that is not retrying.
+	status database.DeliveryStatus
+
+	// inlineBody mirrors a Task built for a body under
+	// MaxInlineBodySize. When false the engine reads the body back
+	// from the stored row.
+	inlineBody bool
+
+	attemptNum int
+
+	process func(
+		ctx context.Context, e *delivery.Engine, task *delivery.Task,
+	)
+}
+
+// run delivers one event through the named path and returns the
+// Slack message text the sink received.
+func (c tsCase) run(t *testing.T) (iSetup, database.Delivery, string) {
+	t.Helper()
+
+	s := newISetup(t)
+	sink := newTSSink(t)
+
+	cfg := tsSlackConfig(t, sink.URL)
+	target := tsSeedTarget(t, s.MainDB, s.WebhookID, cfg)
+	event := tsSeedEvent(t, s.WebhookDB, s.WebhookID)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, target.ID, c.status,
+	)
+
+	var body *string
+
+	if c.inlineBody {
+		bodyStr := event.Body
+		body = &bodyStr
+	}
+
+	task := tsTask(
+		d, event, s.WebhookID, target, c.attemptNum, body,
+	)
+
+	c.process(context.TODO(), s.Engine, &task)
+
+	return s, d, sink.text(t)
+}
+
+// TestSlackFirstAttemptCarriesEventTimestamp covers the path an
+// event takes on its first delivery: the task comes from the
+// receiver and the engine reconstructs the event from it.
+func TestSlackFirstAttemptCarriesEventTimestamp(t *testing.T) {
+	t.Parallel()
+
+	s, d, text := tsCase{
+		status:     database.DeliveryStatusPending,
+		inlineBody: true,
+		attemptNum: 1,
+		process: func(
+			ctx context.Context,
+			e *delivery.Engine,
+			task *delivery.Task,
+		) {
+			e.ExportProcessNewTask(ctx, task)
+		},
+	}.run(t)
+
+	tsAssertRealTimestamp(t, text)
+
+	iAssertStatus(t, s.WebhookDB, d.ID,
+		database.DeliveryStatusDelivered,
+	)
+}
+
+// TestSlackFirstAttemptLargeBodyCarriesEventTimestamp covers the
+// first-attempt path for an event whose body exceeded
+// MaxInlineBodySize, so the task carries no body and the engine
+// reads it back from the stored row.
+func TestSlackFirstAttemptLargeBodyCarriesEventTimestamp(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	_, _, text := tsCase{
+		status:     database.DeliveryStatusPending,
+		inlineBody: false,
+		attemptNum: 1,
+		process: func(
+			ctx context.Context,
+			e *delivery.Engine,
+			task *delivery.Task,
+		) {
+			e.ExportProcessNewTask(ctx, task)
+		},
+	}.run(t)
+
+	tsAssertRealTimestamp(t, text)
+}
+
+// TestSlackRetryCarriesEventTimestamp covers the retry path, which
+// reconstructs the event from the same task the first attempt used.
+func TestSlackRetryCarriesEventTimestamp(t *testing.T) {
+	t.Parallel()
+
+	s, d, text := tsCase{
+		status:     database.DeliveryStatusRetrying,
+		inlineBody: true,
+		attemptNum: 2,
+		process: func(
+			ctx context.Context,
+			e *delivery.Engine,
+			task *delivery.Task,
+		) {
+			e.ExportProcessRetryTask(ctx, task)
+		},
+	}.run(t)
+
+	tsAssertRealTimestamp(t, text)
+
+	iAssertStatus(t, s.WebhookDB, d.ID,
+		database.DeliveryStatusDelivered,
+	)
+}
+
+// TestFormatSlackMessageOverTaskReconstructedEvent asserts on the
+// formatted message directly, over the event the delivery paths
+// reconstruct from a Task. It is the unit-level guard under the
+// end-to-end tests: revert the CreatedAt population in hydrateEvent
+// and this fails on the zero timestamp.
+func TestFormatSlackMessageOverTaskReconstructedEvent(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	s := newISetup(t)
+
+	cfg := tsSlackConfig(t, tsUndeliverableHook)
+	target := tsSeedTarget(t, s.MainDB, s.WebhookID, cfg)
+	event := tsSeedEvent(t, s.WebhookDB, s.WebhookID)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, target.ID,
+		database.DeliveryStatusPending,
+	)
+
+	bodyStr := event.Body
+	task := tsTask(d, event, s.WebhookID, target, 1, &bodyStr)
+
+	rebuilt, err := s.Engine.ExportEventForTask(
+		s.WebhookDB, &task,
+	)
+	require.NoError(t, err)
+	assert.False(t, rebuilt.CreatedAt.IsZero(),
+		"reconstructed event carries the zero time",
+	)
+	assert.Equal(t,
+		tsEventCreatedAt().UTC(), rebuilt.CreatedAt.UTC(),
+	)
+
+	tsAssertRealTimestamp(
+		t, delivery.FormatSlackMessage(&rebuilt),
+	)
+}
+
+// TestFormatSlackMessageZeroTimestamp asserts the rendering choice
+// directly, without going through the engine: a zero CreatedAt (the
+// shape a reaped-row fallback produces) renders as "unknown" rather
+// than the year-1 zero time, while a real CreatedAt still renders as
+// RFC3339.
+func TestFormatSlackMessageZeroTimestamp(t *testing.T) {
+	t.Parallel()
+
+	zeroEvent := database.Event{
+		Method:      http.MethodPost,
+		ContentType: testContentType,
+		Body:        tsEventBody,
+	}
+
+	zeroText := delivery.FormatSlackMessage(&zeroEvent)
+
+	assert.NotContains(t, zeroText, "0001-01-01",
+		"slack message carries the zero-time year",
+	)
+	assert.Contains(t, zeroText, "*Timestamp:* `unknown`",
+		"slack message does not mark an unset receipt time as unknown",
+	)
+
+	nonZeroEvent := zeroEvent
+	nonZeroEvent.CreatedAt = tsEventCreatedAt()
+
+	nonZeroText := delivery.FormatSlackMessage(&nonZeroEvent)
+
+	assert.Contains(t, nonZeroText,
+		"*Timestamp:* `"+
+			tsEventCreatedAt().UTC().Format(time.RFC3339)+"`",
+		"slack message does not render a real receipt time as RFC3339",
+	)
+}
+
+// TestEventReconstructionSurvivesAReapedRow pins the fallback: an
+// event row reaped by retention while its delivery still holds the
+// body inline is still delivered, with the receipt time unset,
+// rather than dropped.
+func TestEventReconstructionSurvivesAReapedRow(t *testing.T) {
+	t.Parallel()
+
+	s := newISetup(t)
+
+	cfg := tsSlackConfig(t, tsUndeliverableHook)
+	target := tsSeedTarget(t, s.MainDB, s.WebhookID, cfg)
+	event := tsSeedEvent(t, s.WebhookDB, s.WebhookID)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, target.ID,
+		database.DeliveryStatusPending,
+	)
+
+	bodyStr := event.Body
+	task := tsTask(d, event, s.WebhookID, target, 1, &bodyStr)
+
+	require.NoError(t, s.WebhookDB.Unscoped().Delete(
+		&database.Event{}, "id = ?", event.ID,
+	).Error)
+
+	rebuilt, err := s.Engine.ExportEventForTask(
+		s.WebhookDB, &task,
+	)
+	require.NoError(t, err)
+	assert.Equal(t, bodyStr, rebuilt.Body)
+	assert.True(t, rebuilt.CreatedAt.IsZero())
+
+	// A task with no inlined body has nothing left to deliver, so
+	// the same reaped row is an error there.
+	noBody := task
+	noBody.Body = nil
+
+	_, err = s.Engine.ExportEventForTask(s.WebhookDB, &noBody)
+	require.Error(t, err)
+}
@@ -33,6 +33,11 @@ const (
 	// response is written against this number, so a test has to
 	// be able to name it.
 	ExportMaxBodyLog = maxBodyLog
+
+	// ExportPendingSweepMinAge is how long a delivery must sit at
+	// pending before the sweep treats it as stranded. A test has to
+	// name it to age a row past the bound.
+	ExportPendingSweepMinAge = pendingSweepMinAge
 )

 // ExportIsBlockedIP exposes isBlockedIP for testing.
@@ -96,6 +101,19 @@ func (e *Engine) ExportDeliverHTTP(
 	e.httpTarget.Deliver(ctx, webhookDB, d, task, e)
 }

+// ExportDeliverHTTPWithScheduler delivers via the http target, handing
+// any retry to sched instead of the engine, so a test can see the
+// delay each retry is given.
+func (e *Engine) ExportDeliverHTTPWithScheduler(
+	ctx context.Context,
+	webhookDB *gorm.DB,
+	d *database.Delivery,
+	task *Task,
+	sched Scheduler,
+) {
+	e.httpTarget.Deliver(ctx, webhookDB, d, task, sched)
+}
+
 // ExportDeliverDatabase delivers via the database target.
 func (e *Engine) ExportDeliverDatabase(
 	webhookDB *gorm.DB, d *database.Delivery,
@@ -146,6 +164,16 @@ func (e *Engine) ExportProcessRetryTask(
 	e.processRetryTask(ctx, task)
 }

+// ExportEventForTask exposes the event reconstruction the delivery
+// paths run: buildEventFromTask followed by hydrateEvent.
+func (e *Engine) ExportEventForTask(
+	webhookDB *gorm.DB, task *Task,
+) (database.Event, error) {
+	return e.hydrateEvent(
+		webhookDB, buildEventFromTask(task), task,
+	)
+}
+
 // ExportProcessDelivery exposes processDelivery.
 func (e *Engine) ExportProcessDelivery(
 	ctx context.Context,
@@ -164,6 +192,14 @@ func (e *Engine) ExportGetCircuitBreaker(
 	return e.httpTarget.getCircuitBreaker(targetID)
 }

+// ExportSetCircuitBreaker makes cb the http target's circuit breaker
+// for targetID, so a test can use one with a short cooldown.
+func (e *Engine) ExportSetCircuitBreaker(
+	targetID string, cb *CircuitBreaker,
+) {
+	e.httpTarget.circuitBreakers.Store(targetID, cb)
+}
+
 // ExportParseHTTPConfig exposes parseHTTPConfig.
 func (e *Engine) ExportParseHTTPConfig(
 	configJSON string,
@@ -286,6 +322,51 @@ func (e *Engine) ExportWedgeWorker(release <-chan struct{}) {
 	})
 }

+// ExportInflightHeld reports how many deliveries the engine currently
+// owns, so a test can prove ownership is released rather than leaked.
+func (e *Engine) ExportInflightHeld() int {
+	return e.inflight.held()
+}
+
+// ExportRetainDelivery takes the first reference on a delivery, as the
+// queueing side does. It lets a test put a delivery into the state a
+// worker or a full channel would, without running the pool.
+func (e *Engine) ExportRetainDelivery(deliveryID string) bool {
+	return e.inflight.retainIdle(deliveryID)
+}
+
+// ExportRecoverRetryingDeliveries exposes recoverRetryingDeliveries.
+func (e *Engine) ExportRecoverRetryingDeliveries(
+	webhookDB *gorm.DB, webhookID string,
+) {
+	e.recoverRetryingDeliveries(webhookDB, webhookID)
+}
+
+// ExportFailMissingTarget exposes failMissingTarget, so a test can hand
+// it a delivery as a batch read it earlier.
+func (e *Engine) ExportFailMissingTarget(
+	webhookDB *gorm.DB,
+	webhookID string,
+	d *database.Delivery,
+) {
+	e.failMissingTarget(webhookDB, webhookID, d)
+}
+
+// ExportSendRecoveredDeliveries exposes sendRecoveredDeliveries, so a
+// test can hand it a target map that lacks a delivery's target.
+func (e *Engine) ExportSendRecoveredDeliveries(
+	ctx context.Context,
+	webhookDB *gorm.DB,
+	deliveries []database.Delivery,
+	webhookID string,
+	targetMap map[string]database.Target,
+	settled map[string]struct{},
+) {
+	e.sendRecoveredDeliveries(
+		ctx, webhookDB, deliveries, webhookID, targetMap, settled,
+	)
+}
+
 // ExportDeliveryCh returns the delivery channel.
 func (e *Engine) ExportDeliveryCh() chan Task {
 	return e.deliveryCh
@@ -0,0 +1,110 @@
+package delivery
+
+import "sync"
+
+// inflightSet records which deliveries the engine currently owns.
+//
+// A delivery is owned from the moment a task for it is handed to a
+// channel or to a retry timer until the engine has no further plan for
+// it in memory. Restart recovery and both arms of the periodic sweep
+// re-dispatch only deliveries the set does not hold, which is what
+// makes them exact rather than a guess about how long a row has sat at
+// pending.
+//
+// This replaces reasoning from timestamps. A delivery's row says
+// pending from creation until its outcome is written, which covers
+// four different situations — never dispatched, waiting in a channel,
+// being attempted right now, and genuinely stranded — and no column
+// distinguishes them. Only the engine knows which, and it knows
+// exactly. `deliveryChannelSize` is 10000 against 10 workers, so a
+// perfectly healthy delivery can wait far longer than any age bound
+// worth setting before its attempt even begins; an age bound alone
+// re-sends it. See
+// https://git.eeqj.de/sneak/webhooker/issues/256.
+//
+// In-memory state is sufficient because a data directory admits one
+// process: internal/datadir takes an flock on it at startup and a
+// second instance refuses to run. Deliveries owned by a process that
+// died are not in any successor's set, and restart recovery is what
+// picks those up.
+//
+// References are counted rather than held as a plain set because
+// ownership outlives the worker that took it. A target that schedules
+// a retry from inside Deliver adds a reference while the worker still
+// holds one, so the delivery stays owned across the gap between the
+// worker returning and the timer firing — the window in which a sweep
+// would otherwise find the row at retrying and send it again.
+//
+// The zero value is ready to use, and the Engine holds one by value.
+// That is deliberate: an engine built by a constructor that forgot to
+// initialise this would not refuse to re-dispatch anything, and the
+// symptom would be duplicate deliveries rather than a failure anybody
+// notices.
+type inflightSet struct {
+	mu  sync.Mutex
+	ids map[string]int
+}
+
+// retain adds a reference to a delivery the caller already knows the
+// engine owns, so that ownership survives the current holder letting
+// go. It cannot fail.
+func (s *inflightSet) retain(deliveryID string) {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	if s.ids == nil {
+		s.ids = make(map[string]int)
+	}
+
+	s.ids[deliveryID]++
+}
+
+// retainIdle takes the first reference on a delivery, and reports
+// whether it got it. It fails when the engine already owns the
+// delivery, which is what makes two claimants — restart recovery and
+// the sweep run concurrently, or two sweep arms — mutually exclusive
+// rather than merely atomic.
+func (s *inflightSet) retainIdle(deliveryID string) bool {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	if s.ids[deliveryID] > 0 {
+		return false
+	}
+
+	if s.ids == nil {
+		s.ids = make(map[string]int)
+	}
+
+	s.ids[deliveryID] = 1
+
+	return true
+}
+
+// release drops one reference. The delivery becomes eligible for
+// re-dispatch again once the last one goes.
+func (s *inflightSet) release(deliveryID string) {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	n := s.ids[deliveryID] - 1
+	if n <= 0 {
+		delete(s.ids, deliveryID)
+
+		return
+	}
+
+	s.ids[deliveryID] = n
+}
+
+// held reports how many deliveries the engine currently owns. It
+// exists so a test can assert that ownership is released rather than
+// leaked: a reference that is never dropped hides its delivery from
+// every sweep for the life of the process, which is the one way this
+// mechanism can fail silently.
+func (s *inflightSet) held() int {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	return len(s.ids)
+}
@@ -0,0 +1,495 @@
+package delivery_test
+
+import (
+	"context"
+	"path/filepath"
+	"strings"
+	"sync"
+	"testing"
+	"time"
+
+	"github.com/google/uuid"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/webhooker/internal/database"
+	"sneak.berlin/go/webhooker/internal/delivery"
+)
+
+// These tests pin the rule that decides whether a delivery may be
+// handed back to a worker: the engine re-dispatches only what it does
+// not already own. Age alone is not that rule — a healthy delivery
+// waiting in a 10000-deep channel is old and must not be re-sent. See
+// https://git.eeqj.de/sneak/webhooker/issues/256.
+
+// fSweepSetup seeds the main database with the webhook row the sweep
+// enumerates, and returns the setup.
+func fSweepSetup(
+	t *testing.T, targetID, name string,
+) iSetup {
+	t.Helper()
+
+	s := newISetup(t)
+
+	iCreateTarget(t, s.MainDB, targetID,
+		s.WebhookID, name,
+		database.TargetTypeLog, "", 0,
+	)
+
+	require.NoError(t, s.MainDB.Create(&database.Webhook{
+		BaseModel: database.BaseModel{ID: s.WebhookID},
+		UserID:    uuid.New().String(),
+		Name:      name,
+	}).Error)
+
+	return s
+}
+
+// fDrain collects every task the engine has queued.
+//
+// Every caller drives the dispatch paths synchronously and has already
+// waited for them to return, so anything they queued is in the channel
+// by now. The short grace covers nothing but scheduler jitter, and is
+// kept small because one of these tests runs the drain forty times.
+func fDrain(e *delivery.Engine) []delivery.Task {
+	var out []delivery.Task
+
+	for {
+		select {
+		case task := <-e.ExportDeliveryCh():
+			out = append(out, task)
+		case task := <-e.ExportRetryCh():
+			out = append(out, task)
+		case <-time.After(25 * time.Millisecond):
+			return out
+		}
+	}
+}
+
+// TestArchiveHandleIsWAL closes the last gap in the durability
+// evidence: the main and per-webhook tiers each assert their journal
+// mode on a live handle, and the archive tier gets its settings from
+// the same code path but nothing checked the running file.
+func TestArchiveHandleIsWAL(t *testing.T) {
+	t.Parallel()
+
+	w := delivery.NewExportArchiveWriter(
+		filepath.Join(t.TempDir(), "archive-wal.db"),
+		archiveTestLogger(), 0,
+	)
+
+	require.NoError(t, w.Open(0))
+
+	var mode string
+
+	row := w.DB().Raw("pragma journal_mode").Row()
+	require.NoError(t, row.Scan(&mode))
+	assert.Equal(t, "wal", strings.ToLower(mode))
+
+	var busy string
+
+	row = w.DB().Raw("pragma busy_timeout").Row()
+	require.NoError(t, row.Scan(&busy))
+	assert.Equal(t, "10000", busy)
+}
+
+// TestSweepLeavesAQueuedDeliveryAlone is the case the age bound cannot
+// see. The delivery is queued and untouched, so its row is arbitrarily
+// old and still perfectly healthy; only ownership distinguishes it
+// from a stranded one.
+func TestSweepLeavesAQueuedDeliveryAlone(t *testing.T) {
+	t.Parallel()
+
+	targetID := uuid.New().String()
+	s := fSweepSetup(t, targetID, "queued")
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"queued":true}`,
+	)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusPending,
+	)
+	rAgePending(t, s.WebhookDB, d.ID)
+
+	// Queued exactly as the receiver queues it, and never dequeued:
+	// no workers are running in this engine.
+	s.Engine.Notify([]delivery.Task{{
+		DeliveryID: d.ID,
+		EventID:    event.ID,
+		WebhookID:  s.WebhookID,
+		TargetID:   targetID,
+	}})
+
+	require.Equal(t, 1, s.Engine.ExportInflightHeld())
+
+	s.Engine.ExportSweepWebhookRetries(
+		context.Background(), s.WebhookID,
+	)
+
+	tasks := fDrain(s.Engine)
+	assert.Len(
+		t, tasks, 1,
+		"the sweep must not queue a delivery that is "+
+			"already waiting for a worker",
+	)
+}
+
+// TestRecoveryAndSweepDoNotDoubleDispatch drives the two entry points
+// the engine starts concurrently against one aged pending row. Before
+// ownership they both dispatched it.
+func TestRecoveryAndSweepDoNotDoubleDispatch(t *testing.T) {
+	t.Parallel()
+
+	targetID := uuid.New().String()
+	s := fSweepSetup(t, targetID, "racing")
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"racing":true}`,
+	)
+
+	ctx := context.Background()
+
+	for range 40 {
+		d := iSeedDelivery(
+			t, s.WebhookDB, event.ID, targetID,
+			database.DeliveryStatusPending,
+		)
+		rAgePending(t, s.WebhookDB, d.ID)
+
+		var wg sync.WaitGroup
+
+		wg.Go(func() {
+			s.Engine.ExportRecoverPendingDeliveries(
+				ctx, s.WebhookDB, s.WebhookID,
+			)
+		})
+		wg.Go(func() {
+			s.Engine.ExportSweepWebhookRetries(
+				ctx, s.WebhookID,
+			)
+		})
+		wg.Wait()
+
+		tasks := fDrain(s.Engine)
+		require.Len(
+			t, tasks, 1,
+			"delivery %s dispatched %d times",
+			d.ID, len(tasks),
+		)
+
+		// No worker runs in this engine, so the reference the winner
+		// took is never released and earlier iterations' deliveries
+		// stay owned — which is itself the property under test, since
+		// both paths see them on every subsequent pass.
+	}
+}
+
+// TestConcurrentClaimsOfOneDeliveryYieldOneOwner exercises the
+// exclusion directly, rather than arguing it from a SQL predicate.
+func TestConcurrentClaimsOfOneDeliveryYieldOneOwner(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	eng := newISetup(t).Engine
+	deliveryID := uuid.New().String()
+
+	var (
+		wg  sync.WaitGroup
+		mu  sync.Mutex
+		won int
+	)
+
+	for range 64 {
+		wg.Go(func() {
+			if eng.ExportRetainDelivery(deliveryID) {
+				mu.Lock()
+				won++
+				mu.Unlock()
+			}
+		})
+	}
+
+	wg.Wait()
+
+	assert.Equal(t, 1, won)
+	assert.Equal(t, 1, eng.ExportInflightHeld())
+}
+
+// TestOwnershipIsReleasedAfterDelivery guards the other direction: a
+// leaked reference hides a delivery from every sweep for the life of
+// the process.
+func TestOwnershipIsReleasedAfterDelivery(t *testing.T) {
+	t.Parallel()
+
+	s := newISetup(t)
+	targetID := uuid.New().String()
+
+	iCreateTarget(t, s.MainDB, targetID,
+		s.WebhookID, "released",
+		database.TargetTypeLog, "", 0,
+	)
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"released":true}`,
+	)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusPending,
+	)
+
+	s.Engine.ExportStart()
+
+	defer func() {
+		require.NoError(
+			t, s.Engine.ExportStop(context.Background()),
+		)
+	}()
+
+	body := `{"released":true}`
+
+	s.Engine.Notify([]delivery.Task{{
+		DeliveryID:   d.ID,
+		EventID:      event.ID,
+		WebhookID:    s.WebhookID,
+		TargetID:     targetID,
+		TargetName:   "released",
+		TargetType:   database.TargetTypeLog,
+		Body:         &body,
+		EntrypointID: event.EntrypointID,
+	}})
+
+	iWaitForDelivered(t, s.WebhookDB, d.ID)
+
+	assert.Eventually(
+		t,
+		func() bool {
+			return s.Engine.ExportInflightHeld() == 0
+		},
+		2*time.Second, 20*time.Millisecond,
+		"the delivery stayed owned after it was delivered",
+	)
+}
+
+// TestNotifyAfterRecoveryDoesNotSendAgain is the startup race of
+// https://git.eeqj.de/sneak/webhooker/issues/299. The receiver has
+// written a delivery, restart recovery finds it pending, sends it and
+// releases it, and only then does the receiver's Notify for it arrive.
+// Nothing owns the delivery by then, so Notify takes it.
+func TestNotifyAfterRecoveryDoesNotSendAgain(t *testing.T) {
+	t.Parallel()
+
+	targetID := uuid.New().String()
+	s := fSweepSetup(t, targetID, "recovered")
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"recovered":true}`,
+	)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusPending,
+	)
+
+	s.Engine.ExportStart()
+
+	defer func() {
+		require.NoError(
+			t, s.Engine.ExportStop(context.Background()),
+		)
+	}()
+
+	// Restart recovery sends the delivery and lets it go.
+	iWaitForDelivered(t, s.WebhookDB, d.ID)
+	require.Eventually(
+		t,
+		func() bool {
+			return s.Engine.ExportInflightHeld() == 0
+		},
+		5*time.Second, 20*time.Millisecond,
+	)
+
+	body := event.Body
+
+	s.Engine.Notify([]delivery.Task{{
+		DeliveryID:   d.ID,
+		EventID:      event.ID,
+		WebhookID:    s.WebhookID,
+		TargetID:     targetID,
+		TargetName:   "recovered",
+		TargetType:   database.TargetTypeLog,
+		Body:         &body,
+		EntrypointID: event.EntrypointID,
+	}})
+
+	// Notify took the delivery, and a worker releases it once it has
+	// run the task.
+	require.Eventually(
+		t,
+		func() bool {
+			return s.Engine.ExportInflightHeld() == 0
+		},
+		5*time.Second, 20*time.Millisecond,
+	)
+
+	assert.Len(
+		t, iResults(t, s.WebhookDB, d.ID), 1,
+		"the delivery was sent a second time",
+	)
+}
+
+// TestRetryingRecoverySkipsASuccessfulResult is the retrying-side twin
+// of the pending reconcile. A second attempt that reached the receiver
+// and whose status write then failed sits at retrying holding a
+// successful result, and re-sending it is the same duplicate.
+func TestRetryingRecoverySkipsASuccessfulResult(t *testing.T) {
+	t.Parallel()
+
+	targetID := uuid.New().String()
+	s := fSweepSetup(t, targetID, "retry-settled")
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"retry":true}`,
+	)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusRetrying,
+	)
+	rSeedResult(t, s.WebhookDB, d.ID, 1, false)
+	rSeedResult(t, s.WebhookDB, d.ID, 2, true)
+
+	s.Engine.ExportRecoverRetryingDeliveries(
+		s.WebhookDB, s.WebhookID,
+	)
+
+	assert.Empty(
+		t, fDrain(s.Engine),
+		"a retrying delivery holding a successful result "+
+			"must not be sent again",
+	)
+
+	iAssertStatus(
+		t, s.WebhookDB, d.ID,
+		database.DeliveryStatusDelivered,
+	)
+}
+
+// TestRetryingSweepSkipsASuccessfulResult is the same rule on the
+// periodic sweep's retrying arm.
+func TestRetryingSweepSkipsASuccessfulResult(t *testing.T) {
+	t.Parallel()
+
+	targetID := uuid.New().String()
+	s := fSweepSetup(t, targetID, "retry-swept")
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"swept":true}`,
+	)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusRetrying,
+	)
+	rSeedResult(t, s.WebhookDB, d.ID, 1, false)
+	rSeedResult(t, s.WebhookDB, d.ID, 2, true)
+
+	s.Engine.ExportSweepWebhookRetries(
+		context.Background(), s.WebhookID,
+	)
+
+	assert.Empty(t, fDrain(s.Engine))
+
+	iAssertStatus(
+		t, s.WebhookDB, d.ID,
+		database.DeliveryStatusDelivered,
+	)
+
+	var attempts int64
+
+	require.NoError(t, s.WebhookDB.
+		Model(&database.DeliveryResult{}).
+		Where("delivery_id = ?", d.ID).
+		Count(&attempts).Error)
+	assert.Equal(
+		t, int64(2), attempts,
+		"settling must not invent an attempt",
+	)
+}
+
+// TestScheduledRetryIsNotSweptDuringBackoff closes the window between
+// a target scheduling a retry and the timer firing. The row says
+// retrying and nothing is running, which is exactly what an orphaned
+// retry looks like from the database.
+func TestScheduledRetryIsNotSweptDuringBackoff(t *testing.T) {
+	t.Parallel()
+
+	targetID := uuid.New().String()
+	s := fSweepSetup(t, targetID, "backoff")
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"backoff":true}`,
+	)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusRetrying,
+	)
+
+	s.Engine.ExportScheduleRetry(delivery.Task{
+		DeliveryID: d.ID,
+		EventID:    event.ID,
+		WebhookID:  s.WebhookID,
+		TargetID:   targetID,
+		AttemptNum: 2,
+	}, time.Hour)
+
+	require.Equal(t, 1, s.Engine.ExportInflightHeld())
+
+	s.Engine.ExportSweepWebhookRetries(
+		context.Background(), s.WebhookID,
+	)
+
+	assert.Empty(
+		t, fDrain(s.Engine),
+		"the sweep must not duplicate a retry that is "+
+			"already scheduled",
+	)
+}
+
+// TestRedispatchStampsTheRow pins the cadence control: a stranded
+// delivery that has just been handed out is not selected again by the
+// next tick a minute later.
+func TestRedispatchStampsTheRow(t *testing.T) {
+	t.Parallel()
+
+	targetID := uuid.New().String()
+	s := fSweepSetup(t, targetID, "stamped")
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"stamped":true}`,
+	)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusPending,
+	)
+	rAgePending(t, s.WebhookDB, d.ID)
+
+	ctx := context.Background()
+
+	s.Engine.ExportSweepWebhookRetries(ctx, s.WebhookID)
+	require.Len(t, fDrain(s.Engine), 1)
+
+	var row database.Delivery
+
+	require.NoError(t, s.WebhookDB.
+		First(&row, "id = ?", d.ID).Error)
+	assert.WithinDuration(
+		t, time.Now(), row.UpdatedAt, time.Minute,
+		"a re-dispatched delivery must be stamped so the "+
+			"next tick does not select it again",
+	)
+}
@@ -229,6 +229,13 @@ func mExhaustRetries(t *testing.T, s iSetup) {
 	body := event.Body
 	cfg := iHTTPConfig(ts.URL)

+	// The retry below is only run if its target still exists; see
+	// https://git.eeqj.de/sneak/webhooker/issues/107.
+	iCreateTarget(
+		t, s.MainDB, targetID, s.WebhookID, "metrics-fail",
+		database.TargetTypeHTTP, cfg, 2,
+	)
+
 	first := iTask(
 		d, event, s.WebhookID, targetID,
 		"metrics-fail", cfg, 2, 1, &body,
@@ -289,6 +296,13 @@ func TestDeliveryMetrics_CircuitBreakerGauge(t *testing.T) {
 	// rather than the budget is what stops the delivery.
 	maxRetries := delivery.ExportDefaultFailureThreshold + 5

+	// The retries below are only run if their target still exists;
+	// see https://git.eeqj.de/sneak/webhooker/issues/107.
+	iCreateTarget(
+		t, s.MainDB, targetID, s.WebhookID, "metrics-trip",
+		database.TargetTypeHTTP, cfg, maxRetries,
+	)
+
 	first := iTask(
 		d, event, s.WebhookID, targetID,
 		"metrics-trip", cfg, maxRetries, 1, &body,
@@ -353,6 +367,11 @@ func TestDeliveryMetrics_BreakerBlockedIsNotAnAttempt(
 	cfg := iHTTPConfig(ts.URL)
 	maxRetries := delivery.ExportDefaultFailureThreshold + 5

+	iCreateTarget(
+		t, s.MainDB, targetID, s.WebhookID, "metrics-blocked",
+		database.TargetTypeHTTP, cfg, maxRetries,
+	)
+
 	first := iTask(
 		d, event, s.WebhookID, targetID,
 		"metrics-blocked", cfg, maxRetries, 1, &body,
@@ -393,9 +412,10 @@ func TestDeliveryMetrics_BreakerBlockedIsNotAnAttempt(

 	s.Engine.ExportProcessRetryTask(context.TODO(), &blocked)

-	// The breaker refused it: rescheduled, so the retry counter
-	// moved, but nothing was attempted or timed.
-	assert.InDelta(t, retriesBefore+1,
+	// The breaker refused it: rescheduled without rewriting the
+	// retrying status it already had, so the retry counter did not
+	// move, and nothing was attempted or timed.
+	assert.InDelta(t, retriesBefore,
 		mCounter(t, reg, mRetries, mTypeHTTP), 0)
 	assert.InDelta(t, threshold,
 		mCounter(t, reg, mAttempts, mTypeHTTP), 0)
@@ -3,8 +3,6 @@ package delivery_test
 import (
 	"bytes"
 	"context"
-	"database/sql"
-	"fmt"
 	"log/slog"
 	"net/http"
 	"path/filepath"
@@ -54,12 +52,10 @@ func (q *qdSyncBuf) String() string {
 func qdMainDB(t *testing.T, log *slog.Logger) *gorm.DB {
 	t.Helper()

-	dsn := fmt.Sprintf(
-		"file:%s?cache=shared&mode=rwc",
+	sqlDB, err := database.OpenSQLite(
 		filepath.Join(t.TempDir(), "main-gormlog.db"),
+		database.SQLiteModeCreate,
 	)
-
-	sqlDB, err := sql.Open("sqlite", dsn)
 	require.NoError(t, err)

 	t.Cleanup(func() { _ = sqlDB.Close() })
@@ -0,0 +1,378 @@
+package delivery_test
+
+import (
+	"context"
+	"net/http"
+	"net/http/httptest"
+	"sync/atomic"
+	"testing"
+	"time"
+
+	"github.com/google/uuid"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"gorm.io/gorm"
+	"sneak.berlin/go/webhooker/internal/database"
+	"sneak.berlin/go/webhooker/internal/delivery"
+)
+
+// These tests cover the delivery half of
+// https://git.eeqj.de/sneak/webhooker/issues/256: a delivery that
+// reached its receiver but whose bookkeeping write failed used to be
+// left at pending and re-sent on the next restart, giving the receiver
+// a second copy while the event log recorded one attempt.
+
+// rSeedResult records a DeliveryResult against a delivery, standing in
+// for the attempt row the send path writes before the status.
+func rSeedResult(
+	t *testing.T,
+	db *gorm.DB,
+	deliveryID string,
+	attemptNum int,
+	success bool,
+) {
+	t.Helper()
+
+	require.NoError(t, db.Create(&database.DeliveryResult{
+		DeliveryID: deliveryID,
+		AttemptNum: attemptNum,
+		Success:    success,
+	}).Error)
+}
+
+// rAgePending backdates a delivery past the sweep's age bound, which is
+// what separates a stranded delivery from one a worker still holds.
+func rAgePending(
+	t *testing.T, db *gorm.DB, deliveryID string,
+) {
+	t.Helper()
+
+	old := time.Now().Add(
+		-2 * delivery.ExportPendingSweepMinAge,
+	)
+
+	require.NoError(t, db.Model(&database.Delivery{}).
+		Where("id = ?", deliveryID).
+		UpdateColumn("updated_at", old).Error)
+}
+
+func TestRecoverySkipsPendingWithSuccessfulResult(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	s := newISetup(t)
+	targetID := uuid.New().String()
+
+	iCreateTarget(t, s.MainDB, targetID,
+		s.WebhookID, "already-delivered",
+		database.TargetTypeLog, "", 0,
+	)
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"delivered":true}`,
+	)
+
+	// The delivery whose send succeeded and whose result row landed:
+	// only the status write failed, so it sits at pending.
+	done := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusPending,
+	)
+	rSeedResult(t, s.WebhookDB, done.ID, 1, true)
+
+	// A delivery that was genuinely never attempted.
+	fresh := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusPending,
+	)
+
+	s.Engine.ExportRecoverPendingDeliveries(
+		context.Background(), s.WebhookDB, s.WebhookID,
+	)
+
+	select {
+	case task := <-s.Engine.ExportDeliveryCh():
+		assert.Equal(
+			t, fresh.ID, task.DeliveryID,
+			"only the unattempted delivery may be re-sent",
+		)
+	case <-time.After(2 * time.Second):
+		t.Fatal("expected the unattempted delivery")
+	}
+
+	select {
+	case task := <-s.Engine.ExportDeliveryCh():
+		t.Fatalf(
+			"re-sent an already delivered delivery: %s",
+			task.DeliveryID,
+		)
+	case <-time.After(200 * time.Millisecond):
+	}
+
+	// It is settled rather than merely skipped: leaving it pending
+	// would strand it again on the next sweep.
+	iAssertStatus(
+		t, s.WebhookDB, done.ID,
+		database.DeliveryStatusDelivered,
+	)
+}
+
+// TestRecoveryContinuesTheAttemptNumbering pins the audit trail: a
+// recovered delivery that already recorded two attempts is re-sent as
+// attempt three, not as attempt one again.
+func TestRecoveryContinuesTheAttemptNumbering(t *testing.T) {
+	t.Parallel()
+
+	s := newISetup(t)
+	targetID := uuid.New().String()
+
+	iCreateTarget(t, s.MainDB, targetID,
+		s.WebhookID, "numbering",
+		database.TargetTypeLog, "", 0,
+	)
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"numbering":true}`,
+	)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusPending,
+	)
+
+	rSeedResult(t, s.WebhookDB, d.ID, 1, false)
+	rSeedResult(t, s.WebhookDB, d.ID, 2, false)
+
+	s.Engine.ExportRecoverPendingDeliveries(
+		context.Background(), s.WebhookDB, s.WebhookID,
+	)
+
+	select {
+	case task := <-s.Engine.ExportDeliveryCh():
+		assert.Equal(t, d.ID, task.DeliveryID)
+		assert.Equal(t, 3, task.AttemptNum)
+	case <-time.After(2 * time.Second):
+		t.Fatal("expected the delivery to be recovered")
+	}
+}
+
+// TestSweepRecoversStrandedPending is the half that removes the
+// restart requirement: a delivery left at pending is picked up by the
+// periodic sweep.
+func TestSweepRecoversStrandedPending(t *testing.T) {
+	t.Parallel()
+
+	targetID := uuid.New().String()
+	s := fSweepSetup(t, targetID, "stranded")
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"stranded":true}`,
+	)
+
+	stranded := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusPending,
+	)
+	rAgePending(t, s.WebhookDB, stranded.ID)
+
+	// A delivery a worker may still be holding: young, and therefore
+	// none of the sweep's business.
+	inFlight := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusPending,
+	)
+
+	s.Engine.ExportSweepWebhookRetries(
+		context.Background(), s.WebhookID,
+	)
+
+	select {
+	case task := <-s.Engine.ExportDeliveryCh():
+		assert.Equal(t, stranded.ID, task.DeliveryID)
+	case <-time.After(2 * time.Second):
+		t.Fatal("expected the stranded delivery")
+	}
+
+	select {
+	case task := <-s.Engine.ExportDeliveryCh():
+		t.Fatalf(
+			"swept an in-flight delivery: %s",
+			task.DeliveryID,
+		)
+	case <-time.After(200 * time.Millisecond):
+	}
+
+	iAssertStatus(
+		t, s.WebhookDB, inFlight.ID,
+		database.DeliveryStatusPending,
+	)
+}
+
+// TestSweepClaimsAStrandedDeliveryOnlyOnce guards the repeat the sweep
+// would otherwise be: the row stays pending for as long as the attempt
+// runs, and a sweep a minute later must not send it a second time.
+func TestSweepClaimsAStrandedDeliveryOnlyOnce(t *testing.T) {
+	t.Parallel()
+
+	targetID := uuid.New().String()
+	s := fSweepSetup(t, targetID, "claimed")
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"claimed":true}`,
+	)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusPending,
+	)
+	rAgePending(t, s.WebhookDB, d.ID)
+
+	ctx := context.Background()
+
+	s.Engine.ExportSweepWebhookRetries(ctx, s.WebhookID)
+
+	select {
+	case task := <-s.Engine.ExportDeliveryCh():
+		assert.Equal(t, d.ID, task.DeliveryID)
+	case <-time.After(2 * time.Second):
+		t.Fatal("expected the stranded delivery")
+	}
+
+	// The delivery is still pending — nothing has run it yet — but
+	// the claim must keep the next sweep off it.
+	iAssertStatus(
+		t, s.WebhookDB, d.ID,
+		database.DeliveryStatusPending,
+	)
+
+	s.Engine.ExportSweepWebhookRetries(ctx, s.WebhookID)
+
+	select {
+	case task := <-s.Engine.ExportDeliveryCh():
+		t.Fatalf(
+			"sent a claimed delivery again: %s",
+			task.DeliveryID,
+		)
+	case <-time.After(200 * time.Millisecond):
+	}
+}
+
+// TestSweepSettlesStrandedPendingWithoutResending is the sweep's own
+// version of the reconcile: a stranded delivery holding a successful
+// result is settled where it stands, and the receiver hears nothing.
+func TestSweepSettlesStrandedPendingWithoutResending(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	targetID := uuid.New().String()
+	s := fSweepSetup(t, targetID, "settled")
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"settled":true}`,
+	)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusPending,
+	)
+	rSeedResult(t, s.WebhookDB, d.ID, 1, true)
+	rAgePending(t, s.WebhookDB, d.ID)
+
+	s.Engine.ExportSweepWebhookRetries(
+		context.Background(), s.WebhookID,
+	)
+
+	select {
+	case task := <-s.Engine.ExportDeliveryCh():
+		t.Fatalf(
+			"re-sent a delivery that already succeeded: %s",
+			task.DeliveryID,
+		)
+	case <-time.After(200 * time.Millisecond):
+	}
+
+	iAssertStatus(
+		t, s.WebhookDB, d.ID,
+		database.DeliveryStatusDelivered,
+	)
+
+	var attempts int64
+
+	require.NoError(t, s.WebhookDB.
+		Model(&database.DeliveryResult{}).
+		Where("delivery_id = ?", d.ID).
+		Count(&attempts).Error)
+	assert.Equal(
+		t, int64(1), attempts,
+		"settling must not invent an attempt",
+	)
+}
+
+// TestFailedResultWriteLeavesDeliveryRecoverable is the rule the
+// targets now follow: a bookkeeping write that fails must not advance
+// the status, because pending and retrying are the states the sweeps
+// recover and delivered is a claim the database refused to record.
+func TestFailedResultWriteLeavesDeliveryRecoverable(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	s := newISetup(t)
+	targetID := uuid.New().String()
+
+	var hits atomic.Int64
+
+	ts := httptest.NewServer(http.HandlerFunc(
+		func(w http.ResponseWriter, _ *http.Request) {
+			hits.Add(1)
+			w.WriteHeader(http.StatusOK)
+		},
+	))
+	defer ts.Close()
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"unwritable":true}`,
+	)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusPending,
+	)
+
+	// Drop the table the attempt row goes in, so the send succeeds
+	// and only the bookkeeping write fails.
+	require.NoError(
+		t,
+		s.WebhookDB.Exec("drop table delivery_results").Error,
+	)
+
+	full := &database.Delivery{
+		EventID:  event.ID,
+		TargetID: targetID,
+		Status:   database.DeliveryStatusPending,
+		Event:    event,
+		Target: database.Target{
+			Name:   "unwritable",
+			Type:   database.TargetTypeHTTP,
+			Config: iHTTPConfig(ts.URL),
+		},
+	}
+	full.ID = d.ID
+
+	s.Engine.ExportDeliverHTTP(
+		context.Background(), s.WebhookDB, full,
+		&delivery.Task{DeliveryID: d.ID, AttemptNum: 1},
+	)
+
+	assert.Equal(
+		t, int64(1), hits.Load(),
+		"the send itself must still happen",
+	)
+
+	iAssertStatus(
+		t, s.WebhookDB, d.ID,
+		database.DeliveryStatusPending,
+	)
+}
@@ -170,7 +170,8 @@ func TestDelivery_CrossOriginRedirectDropsOriginScopedHeaders(
 // Stripping must not fire within the configured origin, or every
 // destination that redirects its own path would lose its
 // credential and start answering 401 — and would lose the inbound
-// signature the receiver verifies.
+// signature header the target endpoint verifies. webhooker's own
+// receiver verifies no signature; it only forwards the header.
 func TestDelivery_SameOriginRedirectKeepsOriginScopedHeaders(
 	t *testing.T,
 ) {
@@ -338,10 +339,11 @@ func TestRedirectPolicy_StopsAtHopCap(t *testing.T) {
 // The set the redirect policy strips is whatever the delivery path
 // actually put on the wire, so a header added to the forward set is
 // covered without a second edit. A header the event never carried
-// is not in the set, and the delivery path's own two are deliberately
-// excluded: Content-Type describes the body, which a 307 carries
-// across hosts, and the inbound User-Agent every real sender supplies
-// is overwritten before the request goes out.
+// is not in the set, and neither is the inbound Content-Type, because
+// it is not forwarded. Two more are deliberately excluded: a
+// Content-Type configured on the target describes the body, which a
+// 307 carries across hosts, and the inbound User-Agent every real
+// sender supplies is overwritten before the request goes out.
 func TestApplyRequestHeaders_ReportsOriginScopedNames(t *testing.T) {
 	t.Parallel()

@@ -370,6 +372,7 @@ func TestApplyRequestHeaders_ReportsOriginScopedNames(t *testing.T) {
 		&delivery.HTTPTargetConfig{
 			Headers: map[string]string{
 				probeHeaderName: probeHeaderValue,
+				"Content-Type":  testContentType,
 			},
 		},
 	)
@@ -377,7 +380,11 @@ func TestApplyRequestHeaders_ReportsOriginScopedNames(t *testing.T) {
 	assert.Equal(t,
 		[]string{probeHeaderName, inboundHeaderName}, names,
 		"both header classes are reported, and only those: "+
-			"Host is never forwarded, Content-Type and "+
-			"User-Agent are the delivery path's own",
+			"Host and the inbound Content-Type are never "+
+			"forwarded, User-Agent is the delivery path's own",
+	)
+	assert.NotContains(t, names, "Content-Type",
+		"a Content-Type configured on the target must survive "+
+			"a cross-origin 307/308 with the body it describes",
 	)
 }
@@ -26,7 +26,7 @@ var (
 		"hostname resolved to no IP addresses",
 	)
 	errBlockedIP = errors.New(
-		"blocked private/reserved IP range",
+		"blocked private, reserved or cloud metadata address",
 	)
 	errBlockedMetadata = errors.New(
 		"blocked link-local or cloud instance metadata " +
@@ -37,9 +37,10 @@ var (
 	)
 )

-// blockedNetworks contains all private/reserved IP ranges
-// that should be blocked to prevent SSRF attacks. An operator
-// can permit specific blocks out of this set with
+// blockedNetworks is the default blocklist: the private and
+// reserved IP ranges, plus the public cloud metadata addresses,
+// that are blocked to prevent SSRF attacks. An operator can
+// permit specific blocks out of this set with
 // ALLOWED_EGRESS_CIDRS; see Guard.
 //
 //nolint:gochecknoglobals // package-level network list is appropriate here
@@ -122,6 +123,8 @@ func init() {
 		"::1/128",
 		"fc00::/7",
 		"fe80::/10",
+		// Azure WireServer, a public address that serves VM credentials.
+		"168.63.129.16/32",
 	})

 	// Every entry is named. The set must not grow or shrink
@@ -216,8 +219,8 @@ func matchesAny(networks []*net.IPNet, ip net.IP) bool {
 }

 // isBlockedIP checks whether an IP address falls within
-// any blocked private/reserved network range, before any
-// operator allowlist is considered.
+// the default blocklist, before any operator allowlist is
+// considered.
 func isBlockedIP(ip net.IP) bool {
 	return matchesAny(blockedNetworks, ip)
 }
@@ -320,7 +323,7 @@ func (g *Guard) allows(ip net.IP) bool {
 //
 //  1. alwaysBlockedNetworks is refused before the allowlist is
 //     consulted, so no configured CIDR reaches link-local or a
-//     cloud instance metadata endpoint.
+//     cloud metadata endpoint at a non-public address.
 //  2. The allowlist is consulted next, so a listed private
 //     network becomes reachable.
 //  3. Everything else keeps the default blocklist's answer.
@@ -390,6 +390,41 @@ func TestGuardAllowlist_PublicUnaffected(t *testing.T) {
 	}
 }

+// TestGuardAllowlist_AzureWireServerReopenable covers Azure's
+// WireServer, a public address that serves VM credentials. The
+// default guard refuses it, but because it is public it sits in
+// the default blocklist rather than the unconditional set, so an
+// operator who lists it can reach it.
+func TestGuardAllowlist_AzureWireServerReopenable(t *testing.T) {
+	t.Parallel()
+
+	const wireServerIP = "168.63.129.16"
+
+	target := "http://" + wireServerIP + "/?comp=versions"
+
+	defaultGuard := delivery.NewTestGuard()
+
+	err := defaultGuard.ValidateTargetURL(context.Background(), target)
+	require.Error(t, err,
+		"WireServer must be refused with no allowlist set",
+	)
+	assert.NotContains(t, err.Error(), metadataRefusalClause,
+		"WireServer must be refused by the default blocklist, "+
+			"which an allowlist can override",
+	)
+
+	assertDialRefused(t, defaultGuard, target)
+
+	listed := delivery.NewTestGuard(
+		netip.MustParsePrefix(wireServerIP + "/32"),
+	)
+
+	assert.NoError(t,
+		listed.ValidateTargetURL(context.Background(), target),
+		"an operator who lists WireServer must be able to reach it",
+	)
+}
+
 // TestGuardCheckIP_BothPathsShareOneDecision asserts that the
 // validator and the dialer are not two policies that happen to
 // agree: both are defined in terms of checkIP, so the exported
@@ -58,12 +58,17 @@ func (t *databaseTarget) Deliver(
 			"error", err,
 		)

-		t.eng.recordResult(
+		recErr := t.eng.recordResult(
 			webhookDB, d, 1, false, 0, "",
 			err.Error(), elapsed.Milliseconds(),
 		)
+		if recErr != nil {
+			t.eng.bookkeepingFailed(d, recErr)

-		t.eng.updateDeliveryStatus(
+			return
+		}
+
+		t.eng.settleStatus(
 			webhookDB, d, d.Target.Type,
 			database.DeliveryStatusFailed,
 		)
@@ -71,12 +76,17 @@ func (t *databaseTarget) Deliver(
 		return
 	}

-	t.eng.recordResult(
+	recErr := t.eng.recordResult(
 		webhookDB, d, 1, true, 0, "", "",
 		elapsed.Milliseconds(),
 	)
+	if recErr != nil {
+		t.eng.bookkeepingFailed(d, recErr)

-	t.eng.updateDeliveryStatus(
+		return
+	}
+
+	t.eng.settleStatus(
 		webhookDB, d, d.Target.Type,
 		database.DeliveryStatusDelivered,
 	)
@@ -267,6 +277,24 @@ func (t *databaseTarget) evict(webhookID string) {
 	)
 }

+// evictAll evicts every cached archive writer, exactly as evict
+// does for one webhook. The engine calls it at shutdown, once its
+// workers have returned. Closing the last handle on an archive
+// moves the contents of its -wal into the .db and removes the
+// -wal, so a clean stop leaves each archive as a single file.
+func (t *databaseTarget) evictAll() {
+	t.mu.Lock()
+
+	writers := t.writers
+	t.writers = nil
+
+	t.mu.Unlock()
+
+	for _, w := range writers {
+		w.evict()
+	}
+}
+
 // sweepWebhook prunes one webhook's archive of rows older than
 // expiry, without requiring a write. It returns nil (nothing to
 // do) when the archive file does not exist, so a sweep never
@@ -1,7 +1,6 @@
 package delivery

 import (
-	"database/sql"
 	"encoding/json"
 	"errors"
 	"fmt"
@@ -12,6 +11,7 @@ import (

 	"gorm.io/driver/sqlite"
 	"gorm.io/gorm"
+	"sneak.berlin/go/webhooker/internal/database"
 	"sneak.berlin/go/webhooker/internal/gormlog"
 )

@@ -30,13 +30,13 @@ const (
 	// path: open the archive file, creating it if missing, so a
 	// first write (or a write after the operator moved the file
 	// away) recreates it.
-	archiveModeCreate = "rwc"
+	archiveModeCreate = database.SQLiteModeCreate

 	// archiveModeExisting is the SQLite URI mode used by the idle
 	// sweep: open read-write but never create. A sweep must never
 	// conjure an empty archive file for a webhook that has a
 	// database target but has never received an event.
-	archiveModeExisting = "rw"
+	archiveModeExisting = database.SQLiteModeExisting
 )

 var (
@@ -273,9 +273,11 @@ func (w *archiveWriter) open(expiry time.Duration) error {
 func (w *archiveWriter) openMode(
 	mode string, expiry time.Duration,
 ) error {
-	dbURL := fmt.Sprintf("file:%s?mode=%s", w.path, mode)
-
-	sqlDB, err := sql.Open("sqlite", dbURL)
+	// Opened through database.OpenSQLite so an archive file carries
+	// the same WAL journaling, busy timeout, immediate-transaction
+	// locking, and pool bounds as every other database file. See
+	// internal/database/sqlite_open.go.
+	sqlDB, err := database.OpenSQLite(w.path, mode)
 	if err != nil {
 		return fmt.Errorf(
 			"opening archive database %s: %w", w.path, err,
@@ -1,6 +1,7 @@
 package delivery_test

 import (
+	"context"
 	"errors"
 	"fmt"
 	"net/http"
@@ -361,3 +362,46 @@ func TestEvictWebhook_LaterDeliveryRecreatesWriter(t *testing.T) {
 		"a later delivery should recreate the writer",
 	)
 }
+
+// TestEngineStop_WriteAfterStopIsRefused proves the engine's stop
+// closes each archive writer the way deleting its webhook does: a
+// write that reaches a writer after the stop is refused, reopens
+// nothing and adds no row.
+func TestEngineStop_WriteAfterStopIsRefused(t *testing.T) {
+	t.Parallel()
+
+	eng, _ := evictTestEngine(t)
+
+	webhookDB := testWebhookDB(t)
+	event := seedEvent(t, webhookDB, `{"archived":true}`)
+	d := seedDatabaseTargetDelivery(t, webhookDB, event, "")
+
+	eng.ExportDeliverDatabase(webhookDB, d)
+
+	w := eng.ExportArchiveWriterFor(event.WebhookID)
+	require.NotNil(t, w)
+	require.True(t, w.HandleOpen())
+
+	require.NoError(t, eng.ExportStop(context.Background()))
+
+	err := w.Write(evictTestRow("ev-after-stop"), 0)
+
+	require.ErrorIs(
+		t, err, delivery.ErrExportArchiveWriterEvicted,
+		"a write after the stop must be refused",
+	)
+	assert.False(
+		t, w.HandleOpen(),
+		"a refused write must not reopen the archive",
+	)
+	assert.False(
+		t, eng.ExportHasArchiveWriter(event.WebhookID),
+		"the stop should empty the registry",
+	)
+
+	count, err := countArchivedRows(w.Path())
+	require.NoError(t, err)
+	assert.Equal(
+		t, int64(1), count, "the refused row must not be written",
+	)
+}
@@ -11,10 +11,11 @@ import (
 	"sneak.berlin/go/webhooker/internal/delivery"
 )

-// Literals these tests repeat, named so that the header name and the
+// Literals these tests repeat, named so that the header names and the
 // keep-forever archive config each have one definition.
 const (
 	headerAuthorization = "Authorization"
+	headerContentType   = "Content-Type"
 	bearerValue         = "Bearer abc"
 	archiveConfigNever  = "{\"expiry\":\"never\"}"
 )
@@ -77,14 +77,19 @@ func (c *httpCore) fireAndForget(
 ) {
 	c.eng.observeAttempt(d.Target.Type, res.elapsed())

-	c.eng.recordResult(
+	err := c.eng.recordResult(
 		webhookDB, d, 1, res.success,
 		res.statusCode, res.respBody, res.errMsg,
 		res.duration,
 	)
+	if err != nil {
+		c.eng.bookkeepingFailed(d, err)
+
+		return
+	}

 	if res.success {
-		c.eng.updateDeliveryStatus(
+		c.eng.settleStatus(
 			webhookDB, d, d.Target.Type,
 			database.DeliveryStatusDelivered,
 		)
@@ -92,7 +97,7 @@ func (c *httpCore) fireAndForget(
 		return
 	}

-	c.eng.updateDeliveryStatus(
+	c.eng.settleStatus(
 		webhookDB, d, d.Target.Type,
 		database.DeliveryStatusFailed,
 	)
@@ -122,16 +127,25 @@ func (c *httpCore) withRetry(

 	c.eng.observeAttempt(d.Target.Type, res.elapsed())

-	c.eng.recordResult(
+	err := c.eng.recordResult(
 		webhookDB, d, attemptNum, res.success,
 		res.statusCode, res.respBody, res.errMsg,
 		res.duration,
 	)
+	if err != nil {
+		// The breaker still learns the outcome: it describes the
+		// target's health, which is unaffected by this database's.
+		c.recordCircuitOutcome(cb, res.success)
+
+		c.eng.bookkeepingFailed(d, err)
+
+		return
+	}

 	if res.success {
 		cb.RecordSuccess()

-		c.eng.updateDeliveryStatus(
+		c.eng.settleStatus(
 			webhookDB, d, d.Target.Type,
 			database.DeliveryStatusDelivered,
 		)
@@ -146,6 +160,20 @@ func (c *httpCore) withRetry(
 	)
 }

+// recordCircuitOutcome feeds one attempt's outcome to the target's
+// circuit breaker.
+func (c *httpCore) recordCircuitOutcome(
+	cb *CircuitBreaker, success bool,
+) {
+	if success {
+		cb.RecordSuccess()
+
+		return
+	}
+
+	cb.RecordFailure()
+}
+
 func (c *httpCore) circuitBreakerBlock(
 	webhookDB *gorm.DB,
 	d *database.Delivery,
@@ -169,10 +197,14 @@ func (c *httpCore) circuitBreakerBlock(
 		"cooldown_remaining", remaining,
 	)

-	c.eng.updateDeliveryStatus(
-		webhookDB, d, d.Target.Type,
-		database.DeliveryStatusRetrying,
-	)
+	// A delivery already at retrying is left as it is, so a task
+	// the breaker keeps turning away writes nothing each time.
+	if d.Status != database.DeliveryStatusRetrying {
+		c.eng.settleStatus(
+			webhookDB, d, d.Target.Type,
+			database.DeliveryStatusRetrying,
+		)
+	}

 	retryTask := *task
 	sched.ScheduleRetry(retryTask, remaining)
@@ -189,7 +221,7 @@ func (c *httpCore) handleRetry(
 	attemptNum int,
 ) {
 	if attemptNum >= maxRetries {
-		c.eng.updateDeliveryStatus(
+		c.eng.settleStatus(
 			webhookDB, d, d.Target.Type,
 			database.DeliveryStatusFailed,
 		)
@@ -197,7 +229,7 @@ func (c *httpCore) handleRetry(
 		return
 	}

-	c.eng.updateDeliveryStatus(
+	c.eng.settleStatus(
 		webhookDB, d, d.Target.Type,
 		database.DeliveryStatusRetrying,
 	)
@@ -332,12 +364,17 @@ func (t *httpTarget) Deliver(
 			"error", err,
 		)

-		t.eng.recordResult(
+		recErr := t.eng.recordResult(
 			webhookDB, d, task.AttemptNum,
 			false, 0, "", err.Error(), 0,
 		)
+		if recErr != nil {
+			t.eng.bookkeepingFailed(d, recErr)

-		t.eng.updateDeliveryStatus(
+			return
+		}
+
+		t.eng.settleStatus(
 			webhookDB, d, d.Target.Type,
 			database.DeliveryStatusFailed,
 		)
@@ -504,6 +541,11 @@ func isForwardableHeader(name string) bool {
 		"Upgrade", "Proxy-Authorization",
 		"Proxy-Connection", "Content-Length":
 		return false
+	case "Content-Type":
+		// applyRequestHeaders sets Content-Type itself. The receiver
+		// already stored this inbound value as the event's
+		// ContentType, so forwarding it too would send it twice.
+		return false
 	default:
 		return true
 	}
@@ -516,6 +558,10 @@ func isForwardableHeader(name string) bool {
 // policy strips exactly that set on a hop that leaves the origin,
 // so the forward set is decided here and only here — a header added
 // to it is covered off-origin without a second edit elsewhere.
+//
+// Content-Type goes out once: a Content-Type configured on the target
+// wins, otherwise the event's ContentType, otherwise none. The inbound
+// Content-Type in the event's headers is never forwarded.
 func applyRequestHeaders(
 	req *http.Request,
 	event *database.Event,
@@ -536,10 +582,10 @@ func applyRequestHeaders(

 	req.Header.Set("User-Agent", "webhooker/1.0")

-	// Content-Type describes the body being sent rather than the
-	// sender, and the delivery path sets it from the event itself.
-	// A 307/308 preserves the body across hosts, so stripping it
-	// would send that body untyped.
+	// A Content-Type configured on the target describes the body
+	// being sent rather than the sender. A 307/308 preserves the
+	// body across hosts, so stripping it would send that body
+	// untyped.
 	delete(originScoped, "Content-Type")

 	// User-Agent is overwritten just above, so an inbound one never
@@ -55,12 +55,17 @@ func (t *logTarget) Deliver(

 	t.eng.observeAttempt(d.Target.Type, elapsed)

-	t.eng.recordResult(
+	err := t.eng.recordResult(
 		webhookDB, d, 1, true, 0, "", "",
 		elapsed.Milliseconds(),
 	)
+	if err != nil {
+		t.eng.bookkeepingFailed(d, err)

-	t.eng.updateDeliveryStatus(
+		return
+	}
+
+	t.eng.settleStatus(
 		webhookDB, d, d.Target.Type,
 		database.DeliveryStatusDelivered,
 	)
@@ -95,12 +95,17 @@ func (t *slackTarget) failConfig(
 	d *database.Delivery,
 	err error,
 ) {
-	t.eng.recordResult(
+	recErr := t.eng.recordResult(
 		webhookDB, d, 1,
 		false, 0, "", err.Error(), 0,
 	)
+	if recErr != nil {
+		t.eng.bookkeepingFailed(d, recErr)

-	t.eng.updateDeliveryStatus(
+		return
+	}
+
+	t.eng.settleStatus(
 		webhookDB, d, d.Target.Type,
 		database.DeliveryStatusFailed,
 	)
@@ -226,10 +231,15 @@ func FormatSlackMessage(
 		event.ContentType,
 	)

+	timestamp := "unknown"
+	if !event.CreatedAt.IsZero() {
+		timestamp = event.CreatedAt.UTC().Format(time.RFC3339)
+	}
+
 	fmt.Fprintf(
 		&b,
 		"*Timestamp:* `%s`\n",
-		event.CreatedAt.UTC().Format(time.RFC3339),
+		timestamp,
 	)

 	fmt.Fprintf(
@@ -0,0 +1,792 @@
+package delivery_test
+
+import (
+	"context"
+	"net/http"
+	"net/http/httptest"
+	"sync/atomic"
+	"testing"
+
+	"github.com/google/uuid"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/webhooker/internal/database"
+	"sneak.berlin/go/webhooker/internal/delivery"
+)
+
+// The two terminal-state gaps of
+// https://git.eeqj.de/sneak/webhooker/issues/107: a delivery failed
+// with nothing in its event log to say why, and a retrying delivery
+// whose target was deleted, which used to keep sending and then never
+// terminalise. Section 4 is the same deleted-target gap for a pending
+// delivery: https://git.eeqj.de/sneak/webhooker/issues/293.
+
+// tUnknownType is a target type no build implements. It stands in for
+// a target whose type was written by a build that knew a type this one
+// does not.
+const tUnknownType = database.TargetType("pubsub")
+
+// tSeedDeletedTarget creates a target, a delivery against it at the
+// given status with one recorded failed attempt, and then deletes the
+// target the way the source page does.
+//
+// It asserts the delete is soft, because that is the whole reason the
+// engine could not tell a deleted target from a target id that never
+// named a row: the surviving row is invisible to a scoped read.
+func tSeedDeletedTarget(
+	t *testing.T,
+	s iSetup,
+	name, url string,
+	status database.DeliveryStatus,
+) string {
+	t.Helper()
+
+	targetID := uuid.New().String()
+
+	iCreateTarget(
+		t, s.MainDB, targetID, s.WebhookID, name,
+		database.TargetTypeHTTP, iHTTPConfig(url), 5,
+	)
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"target":"deleted"}`,
+	)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID, status,
+	)
+
+	iSeedFailedResult(t, s.WebhookDB, d.ID)
+
+	require.NoError(t, s.MainDB.Delete(
+		&database.Target{}, "id = ?", targetID,
+	).Error)
+
+	var scoped, unscoped int64
+
+	require.NoError(t, s.MainDB.
+		Model(&database.Target{}).
+		Where("id = ?", targetID).
+		Count(&scoped).Error)
+
+	require.NoError(t, s.MainDB.Unscoped().
+		Model(&database.Target{}).
+		Where("id = ?", targetID).
+		Count(&unscoped).Error)
+
+	require.Zero(t, scoped,
+		"the deleted target is still visible to a scoped read",
+	)
+	require.Equal(t, int64(1), unscoped,
+		"the delete was hard, so this test proves nothing about "+
+			"the soft-delete case it exists for",
+	)
+
+	return d.ID
+}
+
+// tLastResult returns a delivery's final recorded attempt, asserting
+// the expected number of them.
+func tLastResult(
+	t *testing.T,
+	s iSetup,
+	deliveryID string,
+	want int,
+) database.DeliveryResult {
+	t.Helper()
+
+	results := iResults(t, s.WebhookDB, deliveryID)
+	require.Len(t, results, want)
+
+	return results[want-1]
+}
+
+// --- 1. A failure with nothing recorded ---
+
+func TestProcessDelivery_UnknownTargetType_RecordsWhy(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	s := newISetup(t)
+
+	targetID := uuid.New().String()
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"unknown":"type"}`,
+	)
+
+	seeded := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusPending,
+	)
+
+	target := database.Target{
+		Name:   "mystery",
+		Type:   tUnknownType,
+		Config: iHTTPConfig("http://example.com/hook"),
+	}
+	target.ID = targetID
+
+	d := database.Delivery{
+		EventID:  event.ID,
+		TargetID: targetID,
+		Status:   database.DeliveryStatusPending,
+		Event:    event,
+		Target:   target,
+	}
+	d.ID = seeded.ID
+
+	body := event.Body
+	task := iTask(
+		seeded, event, s.WebhookID, targetID, "mystery",
+		target.Config, 0, 1, &body,
+	)
+	task.TargetType = tUnknownType
+
+	s.Engine.ExportProcessDelivery(
+		context.Background(), s.WebhookDB, &d, &task,
+	)
+
+	iAssertStatus(
+		t, s.WebhookDB, d.ID, database.DeliveryStatusFailed,
+	)
+
+	last := tLastResult(t, s, d.ID, 1)
+
+	assert.False(t, last.Success)
+	assert.Equal(t, 1, last.AttemptNum)
+	assert.Contains(t, last.Error, string(tUnknownType),
+		"the recorded reason does not name the offending type",
+	)
+}
+
+// --- 2. A retrying delivery whose target is gone ---
+
+func TestRecoverSingleRetry_TargetDeleted(t *testing.T) {
+	t.Parallel()
+
+	s := newISetup(t)
+
+	iCreateWebhook(
+		t, s.MainDB, s.WebhookID, "deleted-target-recovery",
+	)
+
+	deliveryID := tSeedDeletedTarget(
+		t, s, "gone-on-recovery", "http://example.com/hook",
+		database.DeliveryStatusRetrying,
+	)
+
+	s.Engine.ExportRecoverWebhookDeliveries(
+		context.Background(), s.WebhookID,
+	)
+
+	iAssertStatus(
+		t, s.WebhookDB, deliveryID,
+		database.DeliveryStatusFailed,
+	)
+
+	last := tLastResult(t, s, deliveryID, 2)
+
+	assert.False(t, last.Success)
+	assert.Equal(t, 2, last.AttemptNum)
+	assert.Contains(t, last.Error, "gone-on-recovery")
+	assert.Contains(t, last.Error, "was deleted")
+
+	assert.Empty(t, s.Engine.ExportRetryCh(),
+		"a delivery whose target is gone was rescheduled",
+	)
+	assert.Zero(t, s.Engine.ExportInflightHeld(),
+		"the terminal path leaked its ownership reference",
+	)
+}
+
+func TestSweepSingleRetry_TargetDeleted(t *testing.T) {
+	t.Parallel()
+
+	s := newISetup(t)
+
+	iCreateWebhook(
+		t, s.MainDB, s.WebhookID, "deleted-target-sweep",
+	)
+
+	deliveryID := tSeedDeletedTarget(
+		t, s, "gone-on-sweep", "http://example.com/hook",
+		database.DeliveryStatusRetrying,
+	)
+
+	// Twice, because the bug was an error the sweep repeated every
+	// minute for the life of the database: the second sweep must
+	// find nothing left to do.
+	s.Engine.ExportSweepWebhookRetries(
+		context.Background(), s.WebhookID,
+	)
+	s.Engine.ExportSweepWebhookRetries(
+		context.Background(), s.WebhookID,
+	)
+
+	iAssertStatus(
+		t, s.WebhookDB, deliveryID,
+		database.DeliveryStatusFailed,
+	)
+
+	last := tLastResult(t, s, deliveryID, 2)
+
+	assert.Contains(t, last.Error, "gone-on-sweep")
+	assert.Contains(t, last.Error, "was deleted")
+
+	assert.Empty(t, s.Engine.ExportRetryCh())
+	assert.Zero(t, s.Engine.ExportInflightHeld())
+}
+
+// TestSweepSingleRetry_TargetNeverExisted covers the other half of the
+// soft-delete distinction: an id with no row at all, deleted or
+// otherwise, must not be reported as something the operator deleted.
+func TestSweepSingleRetry_TargetNeverExisted(t *testing.T) {
+	t.Parallel()
+
+	s := newISetup(t)
+
+	iCreateWebhook(
+		t, s.MainDB, s.WebhookID, "target-never-existed",
+	)
+
+	targetID := uuid.New().String()
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"target":"absent"}`,
+	)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusRetrying,
+	)
+
+	iSeedFailedResult(t, s.WebhookDB, d.ID)
+
+	s.Engine.ExportSweepWebhookRetries(
+		context.Background(), s.WebhookID,
+	)
+
+	iAssertStatus(
+		t, s.WebhookDB, d.ID, database.DeliveryStatusFailed,
+	)
+
+	last := tLastResult(t, s, d.ID, 2)
+
+	assert.Contains(t, last.Error, targetID)
+	assert.Contains(t, last.Error, "no longer exists")
+	assert.NotContains(t, last.Error, "was deleted",
+		"an id that never named a row was reported as a deletion",
+	)
+}
+
+// TestFailMissingTarget_WritesNoTargetRow holds the new terminal path
+// to the same rule as the existing one: no target row, and so no
+// plaintext target config, may be written into the per-webhook event
+// database. See https://git.eeqj.de/sneak/webhooker/issues/206.
+func TestFailMissingTarget_WritesNoTargetRow(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	s := newISetup(t)
+
+	iCreateWebhook(
+		t, s.MainDB, s.WebhookID, "no-target-row-deleted",
+	)
+
+	hookURL := "https://hooks.slack.com/services/T00/B00/x"
+
+	deliveryID := tSeedDeletedTarget(
+		t, s, "credential-bearing", hookURL,
+		database.DeliveryStatusRetrying,
+	)
+
+	s.Engine.ExportSweepWebhookRetries(
+		context.Background(), s.WebhookID,
+	)
+
+	iAssertStatus(
+		t, s.WebhookDB, deliveryID,
+		database.DeliveryStatusFailed,
+	)
+
+	var configs []string
+
+	require.NoError(t, s.WebhookDB.
+		Table("targets").
+		Pluck("config", &configs).Error)
+
+	assert.Empty(t, configs,
+		"the deleted-target terminal path wrote a target row "+
+			"into the per-webhook event database",
+	)
+}
+
+// --- 3. The scheduled retry chain ---
+
+// tRetryChainSetup wires a counting sink and a retrying delivery
+// against a live target pointing at it, and returns the task a
+// scheduled retry would carry — config and all, snapshotted as
+// ScheduleRetry snapshots it.
+func tRetryChainSetup(
+	t *testing.T,
+	s iSetup,
+	name string,
+	hits *atomic.Int64,
+) (delivery.Task, string) {
+	t.Helper()
+
+	ts := httptest.NewServer(http.HandlerFunc(
+		func(w http.ResponseWriter, _ *http.Request) {
+			hits.Add(1)
+			w.WriteHeader(http.StatusOK)
+		},
+	))
+	t.Cleanup(ts.Close)
+
+	iCreateWebhook(t, s.MainDB, s.WebhookID, name)
+
+	targetID := uuid.New().String()
+	cfg := iHTTPConfig(ts.URL)
+
+	iCreateTarget(
+		t, s.MainDB, targetID, s.WebhookID, name,
+		database.TargetTypeHTTP, cfg, 5,
+	)
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"chain":"retry"}`,
+	)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusRetrying,
+	)
+
+	iSeedFailedResult(t, s.WebhookDB, d.ID)
+
+	body := event.Body
+
+	return iTask(
+		d, event, s.WebhookID, targetID, name, cfg, 5, 2, &body,
+	), targetID
+}
+
+// TestProcessRetryTask_TargetDeleted_MakesNoAttempt is the half the
+// deployability audit found worse than filed: terminalising on
+// recovery and sweep alone leaves the already-scheduled timer chain
+// running, and it holds the target's configuration from before the
+// deletion, so it goes on sending to a destination that was removed.
+func TestProcessRetryTask_TargetDeleted_MakesNoAttempt(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	s := newISetup(t)
+
+	var hits atomic.Int64
+
+	task, targetID := tRetryChainSetup(
+		t, s, "gone-mid-chain", &hits,
+	)
+
+	require.NoError(t, s.MainDB.Delete(
+		&database.Target{}, "id = ?", targetID,
+	).Error)
+
+	s.Engine.ExportProcessRetryTask(
+		context.Background(), &task,
+	)
+
+	assert.Zero(t, hits.Load(),
+		"a scheduled retry fired at a target the operator "+
+			"had already deleted",
+	)
+
+	iAssertStatus(
+		t, s.WebhookDB, task.DeliveryID,
+		database.DeliveryStatusFailed,
+	)
+
+	last := tLastResult(t, s, task.DeliveryID, 2)
+
+	assert.False(t, last.Success)
+	assert.Contains(t, last.Error, "was deleted")
+
+	assert.Zero(t, s.Engine.ExportInflightHeld())
+}
+
+// TestProcessRetryTask_TargetPresent_StillDelivers is the guard's
+// mutation check: a liveness check that refused every retry would pass
+// the test above and break every retry there is.
+func TestProcessRetryTask_TargetPresent_StillDelivers(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	s := newISetup(t)
+
+	var hits atomic.Int64
+
+	task, _ := tRetryChainSetup(t, s, "still-there", &hits)
+
+	s.Engine.ExportProcessRetryTask(
+		context.Background(), &task,
+	)
+
+	assert.Equal(t, int64(1), hits.Load())
+
+	iAssertStatus(
+		t, s.WebhookDB, task.DeliveryID,
+		database.DeliveryStatusDelivered,
+	)
+}
+
+// TestProcessRetryTask_TargetUnreadable_StillDelivers pins the other
+// half of the guard: only a target that is confirmed gone stops a
+// retry. A main database that cannot be read is a transient fault, and
+// a guard that abandoned deliveries on one would be a worse bug than
+// the one it fixes.
+func TestProcessRetryTask_TargetUnreadable_StillDelivers(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	s := newISetup(t)
+
+	var hits atomic.Int64
+
+	task, _ := tRetryChainSetup(t, s, "unreadable-main", &hits)
+
+	sqlDB, err := s.MainDB.DB()
+	require.NoError(t, err)
+	require.NoError(t, sqlDB.Close())
+
+	s.Engine.ExportProcessRetryTask(
+		context.Background(), &task,
+	)
+
+	assert.Equal(t, int64(1), hits.Load(),
+		"a retry was abandoned because the main database "+
+			"could not be read, not because its target was gone",
+	)
+
+	iAssertStatus(
+		t, s.WebhookDB, task.DeliveryID,
+		database.DeliveryStatusDelivered,
+	)
+}
+
+// TestRecoverSingleRetry_TargetUnreadable_LeavesDeliveryAlone is the
+// same rule on the recovery path. A read failure that is not
+// "record not found" must leave every retrying delivery of every
+// webhook exactly as it was.
+func TestRecoverSingleRetry_TargetUnreadable_LeavesDeliveryAlone(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	s := newISetup(t)
+
+	iCreateWebhook(
+		t, s.MainDB, s.WebhookID, "unreadable-on-recovery",
+	)
+
+	targetID := uuid.New().String()
+
+	iCreateTarget(
+		t, s.MainDB, targetID, s.WebhookID, "healthy",
+		database.TargetTypeHTTP,
+		iHTTPConfig("http://example.com/hook"), 5,
+	)
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"still":"retrying"}`,
+	)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusRetrying,
+	)
+
+	iSeedFailedResult(t, s.WebhookDB, d.ID)
+
+	sqlDB, err := s.MainDB.DB()
+	require.NoError(t, err)
+	require.NoError(t, sqlDB.Close())
+
+	s.Engine.ExportRecoverRetryingDeliveries(
+		s.WebhookDB, s.WebhookID,
+	)
+
+	iAssertStatus(
+		t, s.WebhookDB, d.ID,
+		database.DeliveryStatusRetrying,
+	)
+
+	assert.Len(t, iResults(t, s.WebhookDB, d.ID), 1,
+		"an unreadable main database produced a terminal "+
+			"failure row",
+	)
+
+	assert.Zero(t, s.Engine.ExportInflightHeld())
+}
+
+// --- 4. A pending delivery whose target is gone ---
+
+func TestRecoverPending_TargetDeleted(t *testing.T) {
+	t.Parallel()
+
+	s := newISetup(t)
+
+	deliveryID := tSeedDeletedTarget(
+		t, s, "gone-while-pending", "http://example.com/hook",
+		database.DeliveryStatusPending,
+	)
+
+	s.Engine.ExportRecoverWebhookDeliveries(
+		context.Background(), s.WebhookID,
+	)
+
+	iAssertStatus(
+		t, s.WebhookDB, deliveryID,
+		database.DeliveryStatusFailed,
+	)
+
+	last := tLastResult(t, s, deliveryID, 2)
+
+	assert.False(t, last.Success)
+	assert.Equal(t, 2, last.AttemptNum)
+	assert.Contains(t, last.Error, "gone-while-pending")
+	assert.Contains(t, last.Error, "was deleted")
+
+	assert.Empty(t, fDrain(s.Engine),
+		"a delivery whose target is gone was sent",
+	)
+	assert.Zero(t, s.Engine.ExportInflightHeld(),
+		"the terminal path leaked its ownership reference",
+	)
+}
+
+// TestRecoverPending_TargetDeleted_LeavesAnOwnedDeliveryAlone: the
+// terminal write takes ownership like every other recovery write, so a
+// delivery the engine still holds is not failed underneath its worker.
+func TestRecoverPending_TargetDeleted_LeavesAnOwnedDeliveryAlone(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	s := newISetup(t)
+
+	deliveryID := tSeedDeletedTarget(
+		t, s, "gone-but-owned", "http://example.com/hook",
+		database.DeliveryStatusPending,
+	)
+
+	require.True(t, s.Engine.ExportRetainDelivery(deliveryID))
+
+	s.Engine.ExportRecoverWebhookDeliveries(
+		context.Background(), s.WebhookID,
+	)
+
+	iAssertStatus(
+		t, s.WebhookDB, deliveryID,
+		database.DeliveryStatusPending,
+	)
+
+	assert.Len(t, iResults(t, s.WebhookDB, deliveryID), 1,
+		"a delivery the engine owns was failed underneath it",
+	)
+}
+
+// TestFailMissingTarget_LeavesASettledDeliveryAlone: the recovery paths
+// read their batch before taking ownership, and a worker may send a
+// delivery and let it go in between. The terminal write goes by the row
+// as it is now, not as the batch read it.
+func TestFailMissingTarget_LeavesASettledDeliveryAlone(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	s := newISetup(t)
+
+	deliveryID := tSeedDeletedTarget(
+		t, s, "gone-after-sending", "http://example.com/hook",
+		database.DeliveryStatusPending,
+	)
+
+	var batch database.Delivery
+
+	require.NoError(t, s.WebhookDB.First(
+		&batch, "id = ?", deliveryID,
+	).Error)
+
+	// A worker settles the delivery after the batch was read.
+	require.NoError(t, s.WebhookDB.Model(&database.Delivery{}).
+		Where("id = ?", deliveryID).
+		Update("status", database.DeliveryStatusDelivered).Error)
+
+	s.Engine.ExportFailMissingTarget(
+		s.WebhookDB, s.WebhookID, &batch,
+	)
+
+	iAssertStatus(
+		t, s.WebhookDB, deliveryID,
+		database.DeliveryStatusDelivered,
+	)
+
+	assert.Len(t, iResults(t, s.WebhookDB, deliveryID), 1,
+		"a delivery settled after the batch read was then failed",
+	)
+	assert.Zero(t, s.Engine.ExportInflightHeld())
+}
+
+// TestSweepPending_TargetDeleted sweeps twice over a batch that also
+// holds a healthy stranded delivery. The one whose target is gone is
+// failed once and then left alone; the healthy one is queued by the
+// first sweep and not again by the second.
+func TestSweepPending_TargetDeleted(t *testing.T) {
+	t.Parallel()
+
+	liveTargetID := uuid.New().String()
+	s := fSweepSetup(t, liveTargetID, "still-there")
+
+	deliveryID := tSeedDeletedTarget(
+		t, s, "gone-on-pending-sweep", "http://example.com/hook",
+		database.DeliveryStatusPending,
+	)
+	rAgePending(t, s.WebhookDB, deliveryID)
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"target":"live"}`,
+	)
+
+	healthy := iSeedDelivery(
+		t, s.WebhookDB, event.ID, liveTargetID,
+		database.DeliveryStatusPending,
+	)
+	rAgePending(t, s.WebhookDB, healthy.ID)
+
+	ctx := context.Background()
+
+	s.Engine.ExportSweepWebhookRetries(ctx, s.WebhookID)
+
+	tasks := fDrain(s.Engine)
+	require.Len(t, tasks, 1,
+		"the first sweep did not queue the healthy delivery",
+	)
+	assert.Equal(t, healthy.ID, tasks[0].DeliveryID)
+
+	s.Engine.ExportSweepWebhookRetries(ctx, s.WebhookID)
+
+	assert.Empty(t, fDrain(s.Engine),
+		"the second sweep queued a delivery again",
+	)
+
+	iAssertStatus(
+		t, s.WebhookDB, deliveryID,
+		database.DeliveryStatusFailed,
+	)
+
+	last := tLastResult(t, s, deliveryID, 2)
+
+	assert.Contains(t, last.Error, "gone-on-pending-sweep")
+	assert.Contains(t, last.Error, "was deleted")
+}
+
+// TestSendRecoveredDeliveries_TargetMissingFromMap: the batch's target
+// map is empty when its query failed, so every delivery in the batch is
+// looked up on its own. A healthy one is sent to the target that lookup
+// finds.
+func TestSendRecoveredDeliveries_TargetMissingFromMap(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	s := newISetup(t)
+
+	targetID := uuid.New().String()
+
+	iCreateTarget(
+		t, s.MainDB, targetID, s.WebhookID, "found-on-lookup",
+		database.TargetTypeLog, "", 0,
+	)
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"map":"empty"}`,
+	)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusPending,
+	)
+
+	s.Engine.ExportSendRecoveredDeliveries(
+		context.Background(), s.WebhookDB,
+		[]database.Delivery{d}, s.WebhookID,
+		map[string]database.Target{}, nil,
+	)
+
+	tasks := fDrain(s.Engine)
+	require.Len(t, tasks, 1,
+		"the healthy delivery was not queued exactly once",
+	)
+	assert.Equal(t, d.ID, tasks[0].DeliveryID)
+	assert.Equal(t, targetID, tasks[0].TargetID)
+	assert.Equal(t, database.TargetTypeLog, tasks[0].TargetType)
+
+	iAssertStatus(
+		t, s.WebhookDB, d.ID,
+		database.DeliveryStatusPending,
+	)
+}
+
+// TestRecoverPending_TargetUnreadable_LeavesDeliveryAlone: a failed
+// read of the main database is not a deleted target. Restart recovery
+// holds every pending delivery of the webhook in one batch, so failing
+// on this would fail all of them.
+func TestRecoverPending_TargetUnreadable_LeavesDeliveryAlone(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	s := newISetup(t)
+
+	targetID := uuid.New().String()
+
+	iCreateTarget(
+		t, s.MainDB, targetID, s.WebhookID, "healthy",
+		database.TargetTypeLog, "", 0,
+	)
+
+	event := iSeedEvent(
+		t, s.WebhookDB, s.WebhookID, `{"still":"pending"}`,
+	)
+
+	d := iSeedDelivery(
+		t, s.WebhookDB, event.ID, targetID,
+		database.DeliveryStatusPending,
+	)
+
+	sqlDB, err := s.MainDB.DB()
+	require.NoError(t, err)
+	require.NoError(t, sqlDB.Close())
+
+	s.Engine.ExportRecoverPendingDeliveries(
+		context.Background(), s.WebhookDB, s.WebhookID,
+	)
+
+	iAssertStatus(
+		t, s.WebhookDB, d.ID,
+		database.DeliveryStatusPending,
+	)
+
+	assert.Empty(t, iResults(t, s.WebhookDB, d.ID),
+		"an unreadable main database produced a terminal "+
+			"failure row",
+	)
+	assert.Empty(t, fDrain(s.Engine))
+	assert.Zero(t, s.Engine.ExportInflightHeld())
+}
@@ -92,11 +92,10 @@ func (h *Handlers) HandleEventBodyDownload() http.HandlerFunc {
 // once per range.
 //
 // One consequence is worth keeping in view: the read finishes
-// before the client is written to, so no read lock is held for
-// the length of a slow download. These per-webhook databases
-// run in SQLite's default journal mode rather than WAL, so a
-// lock held that long would block the receiver from recording
-// new events.
+// before the client is written to, so nothing is held open for
+// the length of a slow download. Under WAL a read no longer
+// blocks the receiver, but it does pin the WAL against
+// checkpointing, and a download can last minutes.
 func (h *Handlers) serveEventBody(
 	w http.ResponseWriter,
 	r *http.Request,
@@ -139,10 +139,10 @@ func (h *Handlers) resubmitEvent(
 	}

 	// Read before the write transaction is opened. The body can be up
-	// to the 1 MB ingest cap, and holding a read of it inside the
-	// transaction would extend how long the per-webhook database is
-	// locked against the receiver, which runs these files in
-	// SQLite's default journal mode rather than WAL.
+	// to the 1 MB ingest cap, and every transaction on these files
+	// takes the write lock at BEGIN (_txlock=immediate, see
+	// internal/database/sqlite_open.go), so reading inside it would
+	// hold that lock against the receiver for the length of the read.
 	src, found, err := loadResubmitSource(
 		webhookDB, webhook.ID, eventID.String(),
 	)
@@ -0,0 +1,283 @@
+package handlers_test
+
+import (
+	"context"
+	"crypto/tls"
+	"net/http"
+	"net/http/httptest"
+	"regexp"
+	"testing"
+
+	"github.com/go-chi/chi"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/webhooker/internal/database"
+	"sneak.berlin/go/webhooker/internal/handlers"
+	"sneak.berlin/go/webhooker/internal/session"
+)
+
+// The only two schemes a rendered entrypoint URL may carry,
+// whatever the request claimed.
+const (
+	schemeHTTPS = "https"
+	schemeHTTP  = "http"
+)
+
+// entrypointURLPattern captures the entrypoint URL the source
+// detail page renders, which is the operator-visible product of
+// BaseURL. Asserting on the extracted string rather than on a
+// substring of the page proves the raw header value cannot reach
+// the scheme by any route.
+var entrypointURLPattern = regexp.MustCompile(
+	`<code id="entrypoint-url-[^"]*"[^>]*>([^<]*)</code>`,
+)
+
+// baseURLFixture is one started app plus the webhook whose
+// entrypoint URL the BaseURL cases read.
+type baseURLFixture struct {
+	handlers *handlers.Handlers
+	session  *session.Session
+	webhook  string
+	path     string
+}
+
+// newBaseURLFixture starts the app and seeds a webhook with one
+// entrypoint.
+func newBaseURLFixture(t *testing.T) *baseURLFixture {
+	t.Helper()
+
+	var (
+		h    *handlers.Handlers
+		sess *session.Session
+		db   *database.Database
+	)
+
+	app := newTestApp(t, &h, &sess, &db)
+	app.RequireStart()
+
+	t.Cleanup(app.RequireStop)
+
+	wh := seedWebhook(t, db)
+	seedEntrypoint(t, db, wh.ID)
+
+	return &baseURLFixture{
+		handlers: h,
+		session:  sess,
+		webhook:  wh.ID,
+		path:     "ep-" + wh.ID,
+	}
+}
+
+// entrypointURL renders the source detail page for the fixture's
+// webhook over a request the caller shapes, and returns the
+// entrypoint URL as an operator would copy it.
+func (f *baseURLFixture) entrypointURL(
+	t *testing.T,
+	host string,
+	shape func(*http.Request),
+) string {
+	t.Helper()
+
+	req := httptest.NewRequestWithContext(
+		context.Background(),
+		http.MethodGet,
+		"/source/"+f.webhook,
+		nil,
+	)
+	req.Host = host
+
+	shape(req)
+
+	for _, c := range authenticatedCookies(
+		t, f.session, deleteTestUserID, deleteTestUsername,
+	) {
+		req.AddCookie(c)
+	}
+
+	rctx := chi.NewRouteContext()
+	rctx.URLParams.Add(paramSourceID, f.webhook)
+	req = req.WithContext(
+		context.WithValue(
+			req.Context(), chi.RouteCtxKey, rctx,
+		),
+	)
+
+	w := httptest.NewRecorder()
+	f.handlers.HandleSourceDetail().ServeHTTP(w, req)
+
+	require.Equal(t, http.StatusOK, w.Code)
+
+	match := entrypointURLPattern.FindStringSubmatch(w.Body.String())
+	require.Len(
+		t, match, 2,
+		"the page must render exactly one entrypoint URL",
+	)
+
+	return match[1]
+}
+
+// forwardedProto returns a request shaper setting
+// X-Forwarded-Proto, or leaving the request alone for "".
+func forwardedProto(value string) func(*http.Request) {
+	return func(r *http.Request) {
+		if value == "" {
+			return
+		}
+
+		r.Header.Set("X-Forwarded-Proto", value)
+	}
+}
+
+// baseURLCase is one X-Forwarded-Proto spelling and the scheme
+// the rendered entrypoint URL owes it.
+type baseURLCase struct {
+	name   string
+	header string
+	scheme string
+	why    string
+}
+
+// baseURLCases enumerate the spellings a proxy really emits. The
+// scheme is only ever http or https: the header value itself is
+// never a scheme, however it is spelled.
+func baseURLCases() []baseURLCase {
+	return []baseURLCase{
+		{
+			name:   "lowercase",
+			header: schemeHTTPS,
+			scheme: schemeHTTPS,
+			why:    "the ordinary spelling",
+		},
+		{
+			name:   "uppercase",
+			header: "HTTPS",
+			scheme: schemeHTTPS,
+			why: "the token is case-insensitive; the scheme " +
+				"in a copyable URL is not",
+		},
+		{
+			name:   "chain with plaintext inner hop",
+			header: "https, http",
+			scheme: schemeHTTPS,
+			why: "a chained proxy appends its hop; the " +
+				"leftmost element faces the client",
+		},
+		{
+			name:   "chain of two TLS hops",
+			header: "https,https",
+			scheme: schemeHTTPS,
+			why:    "appended chain with no space after the comma",
+		},
+		{
+			name:   "trailing space",
+			header: "https ",
+			scheme: schemeHTTPS,
+			why:    "whitespace is not part of the token",
+		},
+		{
+			name:   "plaintext",
+			header: schemeHTTP,
+			scheme: schemeHTTP,
+			why:    "the negative control: the proxy reports plaintext",
+		},
+		{
+			name:   "no header",
+			header: "",
+			scheme: schemeHTTP,
+			why:    "a plaintext request asserting nothing is http",
+		},
+		{
+			name:   "garbage token",
+			header: "javascript:alert(1)//",
+			scheme: schemeHTTP,
+			why: "anything that is not https is not TLS, and " +
+				"the token never becomes the scheme",
+		},
+	}
+}
+
+// TestSourceDetailBaseURL_ForwardedProtoSpellings is the
+// regression test for the entrypoint URL an operator pastes into
+// the sending system: a header spelling that used to land in the
+// scheme verbatim produced a URL no sender could deliver to.
+func TestSourceDetailBaseURL_ForwardedProtoSpellings(t *testing.T) {
+	t.Parallel()
+
+	const host = "hooks.example.com"
+
+	fixture := newBaseURLFixture(t)
+
+	for _, tc := range baseURLCases() {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+
+			assert.Equal(
+				t,
+				tc.scheme+"://"+host+"/webhook/"+fixture.path,
+				fixture.entrypointURL(
+					t, host, forwardedProto(tc.header),
+				),
+				"X-Forwarded-Proto %q: %s", tc.header, tc.why,
+			)
+		})
+	}
+}
+
+// TestSourceDetailBaseURL_DirectTLSBeatsPlaintextHeader pins the
+// precedence the old code had backwards: it let any present
+// header overwrite what the connection itself proved, so a
+// direct-TLS request behind a proxy reporting http rendered an
+// http URL.
+func TestSourceDetailBaseURL_DirectTLSBeatsPlaintextHeader(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	const host = "hooks.example.com"
+
+	fixture := newBaseURLFixture(t)
+
+	got := fixture.entrypointURL(t, host, func(r *http.Request) {
+		r.TLS = &tls.ConnectionState{}
+		r.Header.Set("X-Forwarded-Proto", "http")
+	})
+
+	assert.Equal(
+		t,
+		"https://"+host+"/webhook/"+fixture.path,
+		got,
+		"a connection this process terminated with TLS "+
+			"outranks a header claiming plaintext",
+	)
+}
+
+// TestSourceDetailBaseURL_KeepsHostAuthority pins the host half
+// of the URL: it is taken from the request unchanged, so the
+// deployments that do not sit on port 443 still get a URL that
+// works. Constraining the host would break exactly these.
+func TestSourceDetailBaseURL_KeepsHostAuthority(t *testing.T) {
+	t.Parallel()
+
+	fixture := newBaseURLFixture(t)
+
+	hosts := []string{
+		"hooks.example.com:8443",
+		"[2001:db8::1]:8443",
+		"internal-host",
+	}
+
+	for _, host := range hosts {
+		t.Run(host, func(t *testing.T) {
+			t.Parallel()
+
+			assert.Equal(
+				t,
+				"https://"+host+"/webhook/"+fixture.path,
+				fixture.entrypointURL(
+					t, host, forwardedProto("HTTPS"),
+				),
+				"the authority must survive verbatim, port and all",
+			)
+		})
+	}
+}
@@ -13,6 +13,7 @@ import (
 	"gorm.io/gorm"
 	"sneak.berlin/go/webhooker/internal/database"
 	"sneak.berlin/go/webhooker/internal/delivery"
+	"sneak.berlin/go/webhooker/internal/reqtls"
 )

 // WebhookListItem holds data for the webhook list view.
@@ -427,16 +428,16 @@ func (h *Handlers) renderSourceDetail(
 		}
 	}

-	host := r.Host
-	scheme := "https"
-
-	if r.TLS == nil {
-		scheme = "http"
+	scheme := "http"
+	if reqtls.IsTLS(r) {
+		scheme = "https"
 	}

-	if fwdProto := r.Header.Get("X-Forwarded-Proto"); fwdProto != "" {
-		scheme = fwdProto
-	}
+	// The host is the client's Host header, unvalidated. It is
+	// inert only because source_detail.html renders BaseURL as
+	// text inside a <code> element; putting it in an href or any
+	// other URL context needs it constrained first.
+	baseURL := scheme + "://" + r.Host

 	// The template calls Webhook methods, which take pointer
 	// receivers; html/template cannot address a value stored in a map.
@@ -448,7 +449,7 @@ func (h *Handlers) renderSourceDetail(
 		"Entrypoints": NewEntrypointViews(entrypoints),
 		"Targets":     delivery.NewTargetViews(targets),
 		"Events":      events,
-		"BaseURL":     scheme + "://" + host,
+		"BaseURL":     baseURL,
 	}

 	h.renderTemplate(w, r, "source_detail.html", data)
@@ -300,3 +300,80 @@ func TestEntrypointCopyButtonIsProgressiveEnhancement(t *testing.T) {
 		"the page must render to completion, not abort partway",
 	)
 }
+
+// maxRetriesHelp is the wording both target forms must carry. The
+// delivery core makes max_retries attempts in total, not that many
+// retries on top of a first try (a fresh delivery starts at attempt 1
+// and target_http gives up once the attempt number reaches
+// max_retries), and 0 is special-cased to a single fire-and-forget
+// attempt with no circuit breaker.
+const maxRetriesHelp = "This is the total number of delivery attempts, " +
+	"not retries on top of the first: a value of 3 makes three attempts " +
+	"in all. 0 means a single attempt with no retries and no circuit " +
+	"breaker."
+
+// TestTargetFormMaxRetriesCopyMatchesBehaviour pins the max_retries
+// help text on both the create form (the add-target form on the webhook
+// detail page) and the edit form, so the copy cannot drift back to
+// calling the number a retry count.
+func TestTargetFormMaxRetriesCopyMatchesBehaviour(t *testing.T) {
+	t.Parallel()
+
+	var h *handlers.Handlers
+
+	var sess *session.Session
+
+	app := newTestApp(t, &h, &sess)
+	app.RequireStart()
+
+	t.Cleanup(app.RequireStop)
+
+	webhook := &database.Webhook{Name: "wh", RetentionDays: 14}
+	webhook.ID = testWebhookID
+
+	entrypoint := database.Entrypoint{Path: "abc123"}
+	entrypoint.ID = "ep-1"
+
+	createBody := renderPage(
+		t, h, sess, "source_detail.html", map[string]any{
+			dataKeyWebhook: webhook,
+			"Entrypoints": handlers.NewEntrypointViews(
+				[]database.Entrypoint{entrypoint},
+			),
+			"Targets": delivery.NewTargetViews(nil),
+			"Events":  []database.Event{},
+			"BaseURL": "https://hooks.example.com",
+		},
+	)
+
+	assert.Contains(
+		t, createBody, maxRetriesHelp,
+		"the add-target form must explain max_retries as total attempts",
+	)
+
+	// A slack target exercises the same max_retries field while needing
+	// only Config.URL from the edit template, so the test data stays
+	// minimal. The Target key mirrors the field names the template reads
+	// off the handler's view value.
+	editBody := renderPage(
+		t, h, sess, "target_edit.html", map[string]any{
+			dataKeyWebhook: webhook,
+			"Target": map[string]any{
+				"ID":         "tg-1",
+				"Name":       "t",
+				"Type":       "slack",
+				"Active":     true,
+				"MaxRetries": 3,
+				"Config": map[string]any{
+					"URL": "https://hooks.slack.com/services/x",
+				},
+			},
+			dataKeyError: "",
+		},
+	)
+
+	assert.Contains(
+		t, editBody, maxRetriesHelp,
+		"the target edit form must explain max_retries as total attempts",
+	)
+}
@@ -380,9 +380,8 @@ func csrfTookStrictPath(

 // TestCSRF_ForwardedProtoSpellingsTakeStrictPath runs the header
 // spellings a real proxy emits through the middleware. The environment
-// is dev -- the DEFAULT when WEBHOOKER_ENVIRONMENT is unset -- to pin
-// that the routing is a per-request transport decision and owes
-// nothing to configuration.
+// is set to dev -- the permissive setting -- to pin that the routing is
+// a per-request transport decision and owes nothing to configuration.
 func TestCSRF_ForwardedProtoSpellingsTakeStrictPath(t *testing.T) {
 	t.Parallel()

@@ -133,6 +133,11 @@ func (w *recoverResponseWriter) Unwrap() http.ResponseWriter {
 // what the access log records and the metrics count, and outside the
 // sentryhttp handler, whose Repanic option depends on something
 // further out recovering what it re-raises.
+//
+// Unlike http.Error on its own, it deletes any Set-Cookie the handler
+// set before panicking, because a request that failed must not hand
+// the client a credential; every other header is left to http.Error.
+// See https://git.eeqj.de/sneak/webhooker/issues/193.
 func (s *Middleware) Recoverer() func(http.Handler) http.Handler {
 	return func(next http.Handler) http.Handler {
 		return http.HandlerFunc(func(
@@ -164,6 +169,8 @@ func (s *Middleware) Recoverer() func(http.Handler) http.Handler {
 					return
 				}

+				rw.Header().Del("Set-Cookie")
+
 				http.Error(
 					rw,
 					http.StatusText(
@@ -304,16 +304,44 @@ func TestRecovererRepanicsErrAbortHandler(t *testing.T) {
 	)
 }

+// TestRecovererDropsSetCookieFromTheRecovered500 covers a handler that
+// sets a cookie and a redirect target and then panics before sending
+// anything. A request that failed must not hand the client a
+// credential, so the 500 carries no cookie; Location is left alone.
+func TestRecovererDropsSetCookieFromTheRecovered500(t *testing.T) {
+	t.Parallel()
+
+	probe := newRecovererProbe(
+		t, false,
+		func(w http.ResponseWriter, _ *http.Request) {
+			w.Header().Set("Set-Cookie", "session=x")
+			w.Header().Set("Location", "/after")
+
+			panic(panicMarker)
+		},
+	)
+
+	resp, err := probe.get(t)
+	require.NoError(t, err)
+	require.NoError(t, resp.Body.Close())
+
+	assert.Equal(t, http.StatusInternalServerError, resp.StatusCode)
+	assert.Empty(t, resp.Cookies())
+	assert.Equal(t, "/after", resp.Header.Get("Location"))
+}
+
 // TestRecovererKeepsAnAlreadyCommittedResponse covers a handler that
 // panics after sending its status. The bytes are already on the wire,
-// so a second WriteHeader would change nothing the client sees and
-// would draw net/http's "superfluous response.WriteHeader" report.
+// cookie included, so a second WriteHeader would change nothing the
+// client sees and would draw net/http's "superfluous
+// response.WriteHeader" report.
 func TestRecovererKeepsAnAlreadyCommittedResponse(t *testing.T) {
 	t.Parallel()

 	probe := newRecovererProbe(
 		t, false,
 		func(w http.ResponseWriter, _ *http.Request) {
+			w.Header().Set("Set-Cookie", "session=x")
 			w.WriteHeader(committedStatus)
 			_, _ = w.Write([]byte("partial"))

@@ -331,6 +359,7 @@ func TestRecovererKeepsAnAlreadyCommittedResponse(t *testing.T) {

 	assert.Equal(t, committedStatus, resp.StatusCode)
 	assert.Equal(t, "partial", string(body))
+	assert.Len(t, resp.Cookies(), 1)

 	record := probe.panicRecord(t)
 	assert.Equal(t, panicMarker, record["panic"])
@@ -5,7 +5,7 @@
 // several packages, by hand, and the answers disagreed. The session
 // cookie's Secure attribute was decided at startup from the configured
 // environment while the CSRF cookie's was decided per-request, so a
-// deployment behind a TLS proxy in the default environment emitted one
+// deployment behind a TLS proxy in the dev environment emitted one
 // Secure cookie and one non-Secure cookie on the same response.
 // Everything kept working, which is exactly why nobody noticed.
 //
@@ -0,0 +1,336 @@
+package server_test
+
+import (
+	"context"
+	"net"
+	"net/http"
+	"strconv"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"go.uber.org/fx"
+	"sneak.berlin/go/webhooker/internal/config"
+	"sneak.berlin/go/webhooker/internal/globals"
+	"sneak.berlin/go/webhooker/internal/server"
+)
+
+const (
+	// loopbackV4 is the shipped BIND_ADDRESS default.
+	loopbackV4 = "127.0.0.1"
+
+	// wildcardV4 is the value a container deployment must set,
+	// where a loopback-bound process is unreachable from outside
+	// its network namespace even with a published port.
+	wildcardV4 = "0.0.0.0"
+
+	// unavailableAddr is a TEST-NET-1 address (RFC 5737). It is a
+	// well-formed literal that no host is assigned, so binding it
+	// fails with EADDRNOTAVAIL rather than succeeding somewhere
+	// unexpected.
+	unavailableAddr = "192.0.2.1"
+
+	// listenReadyTimeout bounds the wait for the listener to accept
+	// connections. The bind itself is immediate; this only covers
+	// goroutine scheduling.
+	listenReadyTimeout = 3 * time.Second
+
+	// listenPollInterval is how often the readiness wait retries.
+	listenPollInterval = 10 * time.Millisecond
+
+	// dialTimeout bounds a single connection attempt in these
+	// tests. Everything dialled here is on this host, so a dial
+	// that is not answered immediately is a failure, not slowness.
+	dialTimeout = time.Second
+)
+
+// freePort returns a TCP port that is free on every local address at
+// the moment it returns, by taking one on the wildcard and releasing
+// it. The window between release and re-bind is the standard one
+// every "pick a free port" helper carries.
+func freePort(t *testing.T) int {
+	t.Helper()
+
+	var listenCfg net.ListenConfig
+
+	l, err := listenCfg.Listen(t.Context(), "tcp", "0.0.0.0:0")
+	require.NoError(t, err)
+
+	addr, ok := l.Addr().(*net.TCPAddr)
+	require.True(t, ok, "listener is not TCP")
+	require.NoError(t, l.Close())
+
+	return addr.Port
+}
+
+// otherLocalAddr returns a local IPv4 address that is not
+// loopbackV4, or skips the test when the host has none.
+//
+// The bind-address tests need a second address of this host to stand
+// in for "another interface": what a wildcard bind claims and a
+// loopback bind does not. 127.0.0.2 is that address on Linux, where
+// the whole 127.0.0.0/8 is local; elsewhere an interface address is
+// used instead. Each candidate is proven bindable before it is
+// returned, so a host that offers neither skips rather than fails on
+// something that was never about the code under test.
+func otherLocalAddr(t *testing.T) string {
+	t.Helper()
+
+	candidates := []string{"127.0.0.2"}
+
+	ifaceAddrs, err := net.InterfaceAddrs()
+	require.NoError(t, err)
+
+	for _, a := range ifaceAddrs {
+		ipNet, ok := a.(*net.IPNet)
+		if !ok {
+			continue
+		}
+
+		ip4 := ipNet.IP.To4()
+		if ip4 == nil || ip4.String() == loopbackV4 {
+			continue
+		}
+
+		candidates = append(candidates, ip4.String())
+	}
+
+	var listenCfg net.ListenConfig
+
+	for _, candidate := range candidates {
+		l, listenErr := listenCfg.Listen(
+			t.Context(), "tcp", net.JoinHostPort(candidate, "0"),
+		)
+		if listenErr != nil {
+			continue
+		}
+
+		require.NoError(t, l.Close())
+
+		return candidate
+	}
+
+	t.Skip("host has no second local IPv4 address to bind")
+
+	return ""
+}
+
+// startBoundServer starts the wired app with the given bind address
+// on a free port and returns that port. The app is stopped on
+// cleanup.
+func startBoundServer(t *testing.T, bindAddress string) int {
+	t.Helper()
+
+	port := freePort(t)
+
+	env := newTestEnv(t)
+	env.cfg.BindAddress = bindAddress
+	env.cfg.Port = port
+
+	app := fx.New(
+		fx.NopLogger,
+		fx.Supply(env.log, env.cfg, env.mw, env.hnd),
+		fx.Provide(globals.New, server.New),
+		fx.Invoke(func(*server.Server) {}),
+	)
+
+	startCtx, cancelStart := context.WithTimeout(
+		context.Background(), lifecycleTimeout,
+	)
+	defer cancelStart()
+
+	require.NoError(t, app.Start(startCtx))
+
+	t.Cleanup(func() {
+		stopCtx, cancelStop := context.WithTimeout(
+			context.Background(), lifecycleTimeout,
+		)
+		defer cancelStop()
+
+		require.NoError(t, app.Stop(stopCtx))
+	})
+
+	return port
+}
+
+// dialable reports whether a TCP connection to addr succeeds.
+func dialable(ctx context.Context, addr string) bool {
+	dialer := net.Dialer{Timeout: dialTimeout}
+
+	conn, err := dialer.DialContext(ctx, "tcp", addr)
+	if err != nil {
+		return false
+	}
+
+	_ = conn.Close()
+
+	return true
+}
+
+// requireDialable waits for addr to accept connections, failing the
+// test if it never does.
+func requireDialable(t *testing.T, addr string) {
+	t.Helper()
+
+	deadline := time.Now().Add(listenReadyTimeout)
+	for time.Now().Before(deadline) {
+		if dialable(t.Context(), addr) {
+			return
+		}
+
+		time.Sleep(listenPollInterval)
+	}
+
+	t.Fatalf("nothing accepted connections on %s", addr)
+}
+
+// TestListenAddr pins how BindAddress and Port are rendered into the
+// listen address.
+//
+// The defect this covers was a bare fmt.Sprintf(":%d", port), which
+// binds every interface with no way to say otherwise. The IPv6 rows
+// are here because an unbracketed IPv6 host would produce an address
+// net.Listen rejects, turning a valid configuration into a startup
+// failure.
+func TestListenAddr(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name        string
+		bindAddress string
+		port        int
+		expected    string
+	}{
+		{
+			name:        "loopback default",
+			bindAddress: loopbackV4,
+			port:        8080,
+			expected:    "127.0.0.1:8080",
+		},
+		{
+			name:        "ipv4 wildcard",
+			bindAddress: wildcardV4,
+			port:        8080,
+			expected:    "0.0.0.0:8080",
+		},
+		{
+			name:        "ipv6 wildcard is bracketed",
+			bindAddress: "::",
+			port:        8080,
+			expected:    "[::]:8080",
+		},
+		{
+			name:        "ipv6 literal is bracketed",
+			bindAddress: "2001:db8::5",
+			port:        9001,
+			expected:    "[2001:db8::5]:9001",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			assert.Equal(t, tt.expected, server.ListenAddrForTest(
+				&config.Config{
+					BindAddress: tt.bindAddress,
+					Port:        tt.port,
+				},
+			))
+		})
+	}
+}
+
+// TestBindAddress_LoopbackIsNotOnOtherAddresses proves the fix end to
+// end: with BIND_ADDRESS at its loopback default, the cleartext
+// listener answers on loopback and has not claimed any other address
+// of this host.
+//
+// The second address is proven free by binding it on the same port
+// while the server runs. That is the assertion that fails against the
+// old wildcard bind — a wildcard listener owns the port on every
+// address, so this bind would return EADDRINUSE. Dialling from
+// another machine is what the operator cares about, and this is the
+// in-process form of it: the socket the remote host would connect to
+// does not exist.
+func TestBindAddress_LoopbackIsNotOnOtherAddresses(t *testing.T) {
+	t.Parallel()
+
+	other := otherLocalAddr(t)
+	port := startBoundServer(t, loopbackV4)
+
+	// Positive control: the service really is up and serving.
+	requireDialable(t, net.JoinHostPort(loopbackV4, strconv.Itoa(port)))
+
+	var listenCfg net.ListenConfig
+
+	l, err := listenCfg.Listen(
+		t.Context(), "tcp",
+		net.JoinHostPort(other, strconv.Itoa(port)),
+	)
+	require.NoError(
+		t, err,
+		"port %d on %s is taken while bound to %s: the listener "+
+			"claimed more than its configured address",
+		port, other, loopbackV4,
+	)
+
+	require.NoError(t, l.Close())
+}
+
+// TestBindAddress_WildcardReachesOtherAddresses is the counterpart:
+// the value a container deployment sets does reach the addresses the
+// default withholds. Without this, a loopback-only bind would pass
+// the test above by never listening at all.
+func TestBindAddress_WildcardReachesOtherAddresses(t *testing.T) {
+	t.Parallel()
+
+	other := otherLocalAddr(t)
+	port := startBoundServer(t, wildcardV4)
+
+	requireDialable(t, net.JoinHostPort(other, strconv.Itoa(port)))
+}
+
+// TestBindAddress_ServesRequestsOnConfiguredAddress proves the bound
+// listener serves the application rather than merely accepting TCP,
+// so a bind address that is honoured cannot be mistaken for one that
+// is honoured and broken.
+func TestBindAddress_ServesRequestsOnConfiguredAddress(t *testing.T) {
+	t.Parallel()
+
+	port := startBoundServer(t, loopbackV4)
+	addr := net.JoinHostPort(loopbackV4, strconv.Itoa(port))
+	requireDialable(t, addr)
+
+	req, err := http.NewRequestWithContext(
+		t.Context(), http.MethodGet,
+		"http://"+addr+"/.well-known/healthcheck", nil,
+	)
+	require.NoError(t, err)
+
+	client := &http.Client{Timeout: dialTimeout}
+
+	resp, err := client.Do(req)
+	require.NoError(t, err)
+
+	defer func() { _ = resp.Body.Close() }()
+
+	assert.Equal(t, http.StatusOK, resp.StatusCode)
+}
+
+// TestBindAddress_UnavailableAddressShutsDownTheApp covers the half
+// of the fail-loud rule that configuration parsing cannot reach. A
+// syntactically valid address that is not assigned to this host
+// parses fine and fails at bind time, after fx has already reported
+// RUNNING. It must end the process non-zero rather than leave it
+// alive with nothing listening.
+func TestBindAddress_UnavailableAddressShutsDownTheApp(t *testing.T) {
+	t.Parallel()
+
+	env := newTestEnv(t)
+	env.cfg.BindAddress = unavailableAddr
+	env.cfg.Port = freePort(t)
+
+	requireListenFailureExit(t, env)
+}
@@ -0,0 +1,91 @@
+package server_test
+
+import (
+	"context"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+	"go.uber.org/fx"
+	"sneak.berlin/go/webhooker/internal/globals"
+	"sneak.berlin/go/webhooker/internal/server"
+)
+
+// earlyStopIterations is how many start/stop cycles the race test
+// runs. The window it aims at is the gap between the OnStart hook
+// returning and the serving goroutine reaching its first field
+// access, which is microseconds wide. The race detector reports an
+// unsynchronised pair whenever it observes one, but it has to observe
+// one, so a single cycle can miss purely on scheduling. Repetition
+// makes the observation reliable; the collaborators are built once,
+// so the cycles themselves are cheap.
+const earlyStopIterations = 25
+
+// TestEarlyShutdown_NoPanicAndNoRace stops the application
+// immediately after starting it, before the serving goroutine has
+// necessarily run at all.
+//
+// Two defects live in that window. The OnStart hook returns as soon
+// as it has spawned the serving goroutine, so fx runs the stop
+// sequence against a Server whose serving goroutine may not have
+// executed a single line. cleanShutdown called Shutdown on an
+// httpServer that goroutine was supposed to assign, which was a nil
+// dereference on an early SIGTERM; and it read httpServer and
+// sentryEnabled with nothing ordering those reads against the
+// goroutine's writes, which is a data race that only surfaces once
+// something both starts and stops the server. Nothing did before this
+// test: the listen-failure test never binds, and the router tests
+// bypass the lifecycle entirely.
+//
+// httpServer is now built in New, on the constructing goroutine, so
+// it is written before any hook exists and can never be nil.
+// sentryEnabled is atomic. This test is what catches either one
+// coming back — under -race, which is how the suite runs.
+func TestEarlyShutdown_NoPanicAndNoRace(t *testing.T) {
+	t.Parallel()
+
+	// Built once: the collaborators are not what is under test, and
+	// standing up a database per iteration would make repetition too
+	// expensive to be worth having.
+	env := newTestEnv(t)
+	env.cfg.BindAddress = loopbackV4
+
+	for range earlyStopIterations {
+		requireStartStopIsClean(t, env)
+	}
+}
+
+// requireStartStopIsClean runs one start/stop cycle with no wait in
+// between, failing the test if either half errors.
+//
+// Each cycle gets a fresh fx app, so the Server under test is
+// constructed anew every time — that construction is where the
+// httpServer write now happens, and reusing one Server would test it
+// only once.
+func requireStartStopIsClean(t *testing.T, env *testEnv) {
+	t.Helper()
+
+	env.cfg.Port = freePort(t)
+
+	app := fx.New(
+		fx.NopLogger,
+		fx.Supply(env.log, env.cfg, env.mw, env.hnd),
+		fx.Provide(globals.New, server.New),
+		fx.Invoke(func(*server.Server) {}),
+	)
+
+	startCtx, cancelStart := context.WithTimeout(
+		context.Background(), lifecycleTimeout,
+	)
+	defer cancelStart()
+
+	require.NoError(t, app.Start(startCtx))
+
+	// No sleep and no readiness wait: stopping while the serving
+	// goroutine is still in flight is the whole point.
+	stopCtx, cancelStop := context.WithTimeout(
+		context.Background(), lifecycleTimeout,
+	)
+	defer cancelStop()
+
+	require.NoError(t, app.Stop(stopCtx))
+}
@@ -55,6 +55,16 @@ func NewRouterForTest(
 	return s.router
 }

+// ListenAddrForTest exposes the address the HTTP listener binds for
+// a given Config, so the rendering of host and port — IPv6
+// bracketing above all — can be pinned without standing up a
+// listener.
+func ListenAddrForTest(cfg *config.Config) string {
+	s := &Server{params: ServerParams{Config: cfg}}
+
+	return s.listenAddr()
+}
+
 // ProbePattern is the route NewRouterWithProbeForTest adds to the
 // production route tree.
 const ProbePattern = "/probe"
@@ -80,12 +90,12 @@ func NewRouterWithProbeForTest(
 	probe http.HandlerFunc,
 ) http.Handler {
 	s := &Server{
-		log:           log,
-		mw:            mw,
-		h:             h,
-		params:        ServerParams{Config: cfg},
-		sentryEnabled: sentryEnabled,
+		log:    log,
+		mw:     mw,
+		h:      h,
+		params: ServerParams{Config: cfg},
 	}
+	s.sentryEnabled.Store(sentryEnabled)
 	s.SetupRoutes()
 	s.router.Handle(ProbePattern, probe)

@@ -2,8 +2,9 @@ package server

 import (
 	"errors"
-	"fmt"
+	"net"
 	"net/http"
+	"strconv"
 	"time"
 )

@@ -24,26 +25,52 @@ const (
 	httpMaxHeaderBytes = 1 << 20
 )

-func (s *Server) serveUntilShutdown() {
-	listenAddr := fmt.Sprintf(":%d", s.params.Config.Port)
-	s.httpServer = &http.Server{
-		Addr:           listenAddr,
+// listenAddr renders the address the HTTP listener binds.
+//
+// The host half is always present: an empty host would be the
+// wildcard, and the whole point of BIND_ADDRESS is that binding every
+// interface is a choice the operator makes rather than one the
+// process makes for them. Config guarantees a literal, so
+// JoinHostPort's bracketing is enough to make IPv6 well formed.
+func (s *Server) listenAddr() string {
+	return net.JoinHostPort(
+		s.params.Config.BindAddress,
+		strconv.Itoa(s.params.Config.Port),
+	)
+}
+
+// newHTTPServer builds the HTTP server for this Server's
+// configuration.
+//
+// It is called from New, on the constructing goroutine, rather than
+// from the serving goroutine that used to assign s.httpServer
+// directly. Two goroutines reach that field — the serving goroutine
+// and the fx stop hook, which calls Shutdown on it — with nothing
+// ordering them. Constructing it during New puts the write before
+// every hook fx will later run, which both removes the race and rules
+// out the nil dereference a stop that arrived before the serving
+// goroutine had run would have caused.
+func (s *Server) newHTTPServer() *http.Server {
+	return &http.Server{
+		Addr:           s.listenAddr(),
 		ReadTimeout:    httpReadTimeout,
 		WriteTimeout:   httpWriteTimeout,
 		MaxHeaderBytes: httpMaxHeaderBytes,
 		Handler:        s,
 	}
+}

+func (s *Server) serveUntilShutdown() {
 	// add routes
 	// this does any necessary setup in each handler
 	s.SetupRoutes()

-	s.log.Info("http begin listen", "listenaddr", listenAddr)
+	s.log.Info("http begin listen", "listenaddr", s.httpServer.Addr)

 	err := s.httpServer.ListenAndServe()
 	if err != nil && !errors.Is(err, http.ErrServerClosed) {
 		s.log.Error("listen error", "error", err)
-		s.shutdownOnListenFailure()
+		s.shutdownWithFailure()
 	}
 }

@@ -36,9 +36,9 @@ const lifecycleTimeout = 15 * time.Second
 //
 // The port is occupied by a listener this test holds open, on a
 // kernel-chosen port, so the failure is the real EADDRINUSE the
-// operator hits when a second instance starts. Loopback is enough to
-// collide with the server's wildcard bind: a listening socket on a
-// specific address blocks the wildcard from claiming the same port.
+// operator hits when a second instance starts. The server is pointed
+// at the same loopback address, so the collision is a direct one on
+// the exact address it asks the kernel for.
 func TestListenFailure_ShutsDownTheApp(t *testing.T) {
 	t.Parallel()

@@ -55,11 +55,27 @@ func TestListenFailure_ShutsDownTheApp(t *testing.T) {
 	require.True(t, ok, "listener is not TCP")

 	// The collaborators come from the wired graph rather than stubs,
-	// so the Server under test is the one that ships. Only the port
-	// is test-specific.
+	// so the Server under test is the one that ships. Only the
+	// listen address is test-specific.
 	env := newTestEnv(t)
+	env.cfg.BindAddress = loopbackV4
 	env.cfg.Port = addr.Port

+	requireListenFailureExit(t, env)
+}
+
+// requireListenFailureExit starts the wired app over env and asserts
+// that it gives up on its own with the listen-failure status, then
+// completes its stop sequence.
+//
+// Two different listen failures share it — a port already in use and
+// an address that is not on this host — because what has to hold for
+// both is the same: the failure is discovered after fx has already
+// reported RUNNING, so the only thing that can turn it into a visible
+// exit is the shutdown path under test.
+func requireListenFailureExit(t *testing.T, env *testEnv) {
+	t.Helper()
+
 	app := fx.New(
 		fx.NopLogger,
 		fx.Supply(env.log, env.cfg, env.mw, env.hnd),
@@ -77,7 +93,7 @@ func TestListenFailure_ShutsDownTheApp(t *testing.T) {
 	select {
 	case sig := <-app.Wait():
 		require.Equal(
-			t, server.ListenFailureExitCode, sig.ExitCode,
+			t, server.StartupFailureExitCode, sig.ExitCode,
 			"listen failure must exit non-zero",
 		)
 	case <-time.After(listenFailureDeadline):
@@ -81,7 +81,7 @@ func (s *Server) setupGlobalMiddleware() {
 	// Sentry error reporting (if SENTRY_DSN is set). Repanic is
 	// true so panics still bubble up to the Recoverer middleware
 	// registered immediately above.
-	if s.sentryEnabled {
+	if s.sentryEnabled.Load() {
 		sentryHandler := sentryhttp.New(sentryhttp.Options{
 			Repanic: true,
 		})
@@ -92,11 +92,25 @@ func (s *Server) setupGlobalMiddleware() {
 func (s *Server) setupRoutes() {
 	s.router.Get("/", s.h.HandleIndex())

-	s.router.Mount(
-		"/s",
-		http.StripPrefix("/s", http.FileServer(http.FS(static.Static))),
+	// Static assets answer GET and HEAD only. chi's default 405
+	// carries no Allow header, so this group supplies its own.
+	staticFiles := http.StripPrefix(
+		"/s", http.FileServer(http.FS(static.Static)),
 	)

+	s.router.Route("/s", func(r chi.Router) {
+		r.MethodNotAllowed(func(w http.ResponseWriter, _ *http.Request) {
+			w.Header().Set("Allow", "GET, HEAD")
+			http.Error(
+				w,
+				"Method Not Allowed",
+				http.StatusMethodNotAllowed,
+			)
+		})
+		r.Method(http.MethodGet, "/*", staticFiles)
+		r.Method(http.MethodHead, "/*", staticFiles)
+	})
+
 	s.router.Route("/api/v1", func(_ chi.Router) {
 		// API routes will be added here.
 	})
@@ -7,6 +7,7 @@ import (
 	"net/http/httptest"
 	"net/url"
 	"regexp"
+	"slices"
 	"strconv"
 	"strings"
 	"testing"
@@ -220,9 +221,21 @@ func (e *testEnv) csrfFrom(
 	// out of the markup has to be unescaped before it is submitted.
 	token := html.UnescapeString(match[1])

-	combined := make([]*http.Cookie, 0, len(cookies))
-	combined = append(combined, cookies...)
-	combined = append(combined, w.Result().Cookies()...)
+	// A cookie the page sets replaces the one of the same name, as in
+	// a browser. Sent both, the server would read the first, older one.
+	set := w.Result().Cookies()
+	combined := make([]*http.Cookie, 0, len(cookies)+len(set))
+
+	for _, c := range cookies {
+		replaced := slices.ContainsFunc(set, func(n *http.Cookie) bool {
+			return n.Name == c.Name
+		})
+		if !replaced {
+			combined = append(combined, c)
+		}
+	}
+
+	combined = append(combined, set...)

 	return token, combined
 }
@@ -396,13 +409,15 @@ func (e *testEnv) storedHash(t *testing.T, username string) string {

 // --- /s static group ---

-// TestStaticServesEveryMethod pins what the static mount actually
-// answers. chi's Mount registers the handler for all methods and
-// http.FileServer only special-cases HEAD (by suppressing the body),
-// so a POST or a DELETE to an asset is served the file rather than
-// refused. The README documents this; the test is what keeps the two
-// from drifting.
-func TestStaticServesEveryMethod(t *testing.T) {
+// TestStaticServesOnlyGetAndHead pins the methods the static group
+// answers: GET and HEAD are served the asset, and the other methods
+// chi routes (POST, PUT, DELETE and the rest) are refused with 405
+// and an Allow header naming those two. A method chi does not route,
+// such as PROPFIND, is refused with 405 by the top-level router
+// before it reaches the static group, so it gets no Allow header.
+// The README documents this; the test is what keeps the two from
+// drifting.
+func TestStaticServesOnlyGetAndHead(t *testing.T) {
 	t.Parallel()

 	env := newTestEnv(t)
@@ -417,6 +432,7 @@ func TestStaticServesEveryMethod(t *testing.T) {
 		http.MethodPost,
 		http.MethodPut,
 		http.MethodDelete,
+		"PROPFIND",
 	} {
 		t.Run(method, func(t *testing.T) {
 			t.Parallel()
@@ -428,18 +444,38 @@ func TestStaticServesEveryMethod(t *testing.T) {
 			w := httptest.NewRecorder()
 			env.router.ServeHTTP(w, req)

-			assert.Equal(t, http.StatusOK, w.Code,
-				"static mount answers every method")
-
-			if method == http.MethodHead {
+			switch method {
+			case http.MethodGet:
+				assert.Equal(t, http.StatusOK, w.Code)
+				assert.Equal(t, body, w.Body.Bytes(),
+					"the asset itself is returned")
+			case http.MethodHead:
+				assert.Equal(t, http.StatusOK, w.Code)
 				assert.Empty(t, w.Body.Bytes(),
 					"HEAD must not carry a body")
-
-				return
+			case "PROPFIND":
+				assert.Equal(
+					t, http.StatusMethodNotAllowed, w.Code,
+				)
+				assert.Empty(t, w.Header().Get("Allow"),
+					"chi refuses a method it does not route "+
+						"before the static group runs")
+				assert.NotContains(
+					t, w.Body.String(), string(body),
+					"a refused method must not get the asset",
+				)
+			default:
+				assert.Equal(
+					t, http.StatusMethodNotAllowed, w.Code,
+				)
+				assert.Equal(
+					t, "GET, HEAD", w.Header().Get("Allow"),
+				)
+				assert.NotContains(
+					t, w.Body.String(), string(body),
+					"a refused method must not get the asset",
+				)
 			}
-
-			assert.Equal(t, body, w.Body.Bytes(),
-				"the asset itself is returned")
 		})
 	}
 }
@@ -591,6 +627,59 @@ func TestPagesLogin_CorrectPasswordSurvivesASpentBudget(
 	)
 }

+// TestPagesLogin_CookiesFromAnEarlierDatabase is
+// https://git.eeqj.de/sneak/webhooker/issues/359. A new database
+// brings a new session key, and the operator's browser still holds
+// the session and CSRF cookies signed with the old one. Logging in
+// must work as from a fresh browser and leave cookies the new key
+// accepts.
+func TestPagesLogin_CookiesFromAnEarlierDatabase(t *testing.T) {
+	t.Parallel()
+
+	const (
+		username = "operator"
+		password = "correct-horse-battery-staple"
+	)
+
+	earlier := newTestEnv(t)
+	earlierID, _ := earlier.seedUser(t, username, password)
+	_, stale := earlier.csrfFrom(t, "/pages/login", nil)
+	stale = append(stale, earlier.authCookies(t, earlierID, username)...)
+
+	env := newTestEnv(t)
+	env.seedUser(t, username, password)
+
+	token, cookies := env.csrfFrom(t, "/pages/login", stale)
+
+	form := url.Values{}
+	form.Set("csrf_token", token)
+	form.Set("username", username)
+	form.Set("password", password)
+
+	w := env.post("/pages/login", form, cookies)
+	require.Equal(
+		t, http.StatusSeeOther, w.Code,
+		"a session cookie from another key must not fail the login",
+	)
+
+	// The response deletes the old session cookie and then sets the
+	// new one; a browser keeps the last.
+	var fresh *http.Cookie
+
+	for _, c := range w.Result().Cookies() {
+		if c.Name == session.SessionName {
+			fresh = c
+		}
+	}
+
+	require.NotNil(t, fresh, "login must set a session cookie")
+	assert.Equal(
+		t, "/sources",
+		env.get("/", []*http.Cookie{fresh}).Header().Get("Location"),
+		"the new session cookie must authenticate",
+	)
+}
+
 // --- /user/{username} group ---

 // TestPasswordChange_OversizeBody_RejectedAndPasswordUnchanged
@@ -0,0 +1,99 @@
+package server_test
+
+import (
+	"context"
+	"net"
+	"strconv"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/require"
+	"go.uber.org/fx"
+	"sneak.berlin/go/webhooker/internal/config"
+	"sneak.berlin/go/webhooker/internal/globals"
+	"sneak.berlin/go/webhooker/internal/server"
+)
+
+// TestSentryInitFailure_ShutsDownTheApp pins that error reporting
+// which is configured and cannot be started ends the application
+// instead of serving without it.
+//
+// The measured defect logged `sentry init failure` and kept running,
+// so the deployment served traffic with reporting off while every
+// other signal — SENTRY_DSN still set, the startup summary's own
+// field — said it was on. Nothing later in the process can notice
+// that reports are going nowhere, which is why this exits rather than
+// degrades.
+//
+// The DSN is placed on a hand-built Config, which is the only way to
+// reach this branch at all: loadFromEnv now parses SENTRY_DSN with
+// sentry.NewDsn, the same call sentry.Init makes, so a DSN that
+// survives configuration cannot fail initialisation in the SDK
+// version this pins. The branch stays because that is a property of
+// the SDK's current implementation rather than of its contract.
+func TestSentryInitFailure_ShutsDownTheApp(t *testing.T) {
+	t.Parallel()
+
+	port := freePort(t)
+
+	env := newTestEnvWithConfig(t, &config.Config{
+		DataDir:     t.TempDir(),
+		Environment: config.EnvironmentDev,
+		BindAddress: loopbackV4,
+		Port:        port,
+		SentryDSN:   "not-a-dsn",
+	})
+
+	app := fx.New(
+		fx.NopLogger,
+		fx.Supply(env.log, env.cfg, env.mw, env.hnd),
+		fx.Provide(globals.New, server.New),
+		fx.Invoke(func(*server.Server) {}),
+	)
+
+	startCtx, cancelStart := context.WithTimeout(
+		context.Background(), lifecycleTimeout,
+	)
+	defer cancelStart()
+
+	require.NoError(t, app.Start(startCtx))
+
+	select {
+	case sig := <-app.Wait():
+		require.Equal(
+			t, server.StartupFailureExitCode, sig.ExitCode,
+			"a sentry failure must exit non-zero",
+		)
+	case <-time.After(listenFailureDeadline):
+		t.Fatal("a sentry failure left the app running")
+	}
+
+	// The stop sequence still has to complete: the failure must reach
+	// shutdown through fx rather than around it.
+	stopCtx, cancelStop := context.WithTimeout(
+		context.Background(), lifecycleTimeout,
+	)
+	defer cancelStop()
+
+	require.NoError(t, app.Stop(stopCtx))
+
+	// And it must give up before it listens. A process that bound the
+	// port and then exited would have accepted requests it could not
+	// report on, which is the state under test in miniature.
+	requireBindable(t, port)
+}
+
+// requireBindable asserts that the port is free, which it is only if
+// the server under test never claimed it.
+func requireBindable(t *testing.T, port int) {
+	t.Helper()
+
+	var listenCfg net.ListenConfig
+
+	listener, err := listenCfg.Listen(
+		t.Context(), "tcp",
+		net.JoinHostPort(loopbackV4, strconv.Itoa(port)),
+	)
+	require.NoError(t, err, "the server bound a port it then gave up")
+	require.NoError(t, listener.Close())
+}
@@ -9,6 +9,7 @@ import (
 	"net/http"
 	"os"
 	"os/signal"
+	"sync/atomic"
 	"syscall"
 	"time"

@@ -50,12 +51,13 @@ const (
 	minSentryFlush = 250 * time.Millisecond
 )

-// ListenFailureExitCode is the status the process exits with when the
-// HTTP listener cannot be established, or dies for a reason other
-// than a requested shutdown. It must stay non-zero: systemd
-// `Restart=on-failure` and Docker's restart policies key off it, and a
-// zero exit would read as a deliberate stop.
-const ListenFailureExitCode = 1
+// StartupFailureExitCode is the status the process exits with when
+// the serving goroutine gives up: the HTTP listener cannot be
+// established or dies for a reason other than a requested shutdown, or
+// error reporting is configured and cannot be started. It must stay
+// non-zero: systemd `Restart=on-failure` and Docker's restart policies
+// key off it, and a zero exit would read as a deliberate stop.
+const StartupFailureExitCode = 1

 // SentryFlushBudget reports how long the Sentry flush may run when
 // remaining is the time left on the fx stop context after the HTTP
@@ -88,8 +90,15 @@ type ServerParams struct {
 // Server is the main HTTP server that wires up routes and manages
 // graceful shutdown.
 type Server struct {
-	startupTime   time.Time
-	sentryEnabled bool
+	startupTime time.Time
+
+	// sentryEnabled is written by the serving goroutine, in
+	// enableSentry, and read by the fx stop hook in cleanShutdown.
+	// Nothing orders those two: the OnStart hook returns as soon as
+	// the goroutine is spawned, so a stop can be running while
+	// enableSentry is still deciding. It is atomic to supply the
+	// edge the goroutines do not.
+	sentryEnabled atomic.Bool
 	log           *slog.Logger
 	cancelFunc    context.CancelFunc
 	httpServer    *http.Server
@@ -107,6 +116,7 @@ func New(lc fx.Lifecycle, params ServerParams) (*Server, error) {
 	s.mw = params.Middleware
 	s.h = params.Handlers
 	s.log = params.Logger.Get()
+	s.httpServer = s.newHTTPServer()

 	lc.Append(fx.Hook{
 		OnStart: func(_ context.Context) error {
@@ -126,11 +136,25 @@ func New(lc fx.Lifecycle, params ServerParams) (*Server, error) {
 }

 // Run configures Sentry and starts serving HTTP requests.
+//
+// A Sentry failure ends the application instead of listening. It runs
+// before the listener rather than after it so that the process never
+// binds a port it is about to give up.
 func (s *Server) Run() {
 	s.configure()

 	// logging before sentry, because sentry logs
-	s.enableSentry()
+	err := s.enableSentry()
+	if err != nil {
+		s.log.Error(
+			"SENTRY_DSN is set but error reporting could not be "+
+				"started; refusing to serve with it off",
+			"error", err,
+		)
+		s.shutdownWithFailure()
+
+		return
+	}

 	s.serve()
 }
@@ -141,11 +165,23 @@ func (s *Server) MaintenanceMode() bool {
 	return s.params.Config.MaintenanceMode
 }

-func (s *Server) enableSentry() {
-	s.sentryEnabled = false
+// enableSentry initialises the Sentry SDK when error reporting is
+// configured, and reports the failure when it is configured and cannot
+// be initialised. A DSN that is not set is not a failure: reporting
+// stays off and the server starts normally.
+//
+// There is no fallback to running with reporting off. An operator who
+// set SENTRY_DSN asked for failures to be visible, and serving traffic
+// with reporting quietly off is the one state nothing can ever tell
+// them about — the DSN is still set, so every later signal says it is
+// on. Config already refused a DSN the SDK cannot parse, which is what
+// a typo produces, so reaching this branch means the SDK refused
+// something that parsed: not a condition to guess at either.
+func (s *Server) enableSentry() error {
+	s.sentryEnabled.Store(false)

-	if s.params.Config.SentryDSN == "" {
-		return
+	if !s.params.Config.SentryEnabled() {
+		return nil
 	}

 	err := sentry.Init(sentryClientOptions(
@@ -157,19 +193,19 @@ func (s *Server) enableSentry() {
 		),
 	))
 	if err != nil {
-		s.log.Error("sentry init failure", "error", err)
-		// Don't use fatal since we still want the service to run
-		return
+		return fmt.Errorf("initialising sentry: %w", err)
 	}

 	s.log.Info("sentry error reporting activated")
-	s.sentryEnabled = true
+	s.sentryEnabled.Store(true)
+
+	return nil
 }

 // serve installs the signal watcher, starts the listener and blocks
 // until the server's context is cancelled. The process exit status is
 // fx's to decide — from a signal, or from the code
-// shutdownOnListenFailure hands the Shutdowner — so this reports
+// shutdownWithFailure hands the Shutdowner — so this reports
 // nothing back to its caller.
 func (s *Server) serve() {
 	ctx, cancelFunc := context.WithCancel(context.Background())
@@ -199,20 +235,24 @@ func (s *Server) serve() {
 	// Do not call cleanShutdown() here to avoid double invocation.
 }

-// shutdownOnListenFailure ends the application after the HTTP
-// listener failed. The fx OnStart hook returns as soon as the serving
-// goroutine is spawned, so nothing downstream of it ever learns that
-// the listen failed: fx reports RUNNING and the process sits alive
-// with nothing bound, which is invisible to systemd and Docker
-// restart policies. Asking the Shutdowner to stop the app with a
-// non-zero code is what turns that into a visible failure.
+// shutdownWithFailure ends the application non-zero from the serving
+// goroutine. It is how anything on that goroutine fails fatally: the
+// fx OnStart hook returns as soon as the goroutine is spawned, so
+// nothing downstream of it ever learns that the goroutine gave up. fx
+// reports RUNNING and the process sits alive having done neither what
+// it was asked nor anything visible instead, which systemd and
+// Docker restart policies cannot see. Asking the Shutdowner to stop
+// the app with a non-zero code is what turns that into a visible
+// failure, and it is the whole of "fatal" here — no panic, no
+// os.Exit, and every stop hook still runs.
 //
-// The context cancel that follows only unwinds serve()'s own wait.
-// The shutdown itself runs through fx's normal stop sequence, so the
-// clean-shutdown drain in cleanShutdown is reached unchanged.
-func (s *Server) shutdownOnListenFailure() {
+// The context cancel that follows only unwinds serve()'s own wait,
+// and is skipped before serve has installed one. The shutdown itself
+// runs through fx's normal stop sequence, so the clean-shutdown drain
+// in cleanShutdown is reached unchanged.
+func (s *Server) shutdownWithFailure() {
 	err := s.params.Shutdowner.Shutdown(
-		fx.ExitCode(ListenFailureExitCode),
+		fx.ExitCode(StartupFailureExitCode),
 	)
 	if err != nil {
 		s.log.Error("shutdown request failed", "error", err)
@@ -242,7 +282,7 @@ func (s *Server) cleanShutdown(ctx context.Context) {

 	s.cleanupForExit()

-	if s.sentryEnabled {
+	if s.sentryEnabled.Load() {
 		s.flushSentry(ctx)
 	}
 }
@@ -19,8 +19,8 @@ import (
 )

 // The tests below exercise the securecookie codecs underneath the
-// store and nothing else: Session.Get only decodes, so no server-side
-// expiry check takes part in the result. They exist because
+// store and nothing else: they decode through the store itself, so no
+// server-side expiry check takes part in the result. They exist because
 // NewCookieStore gives its codecs a 30-day max age that assigning
 // store.Options does not override, which would let the codec accept a
 // cookie weeks past the cap the cookie attribute advertises.
@@ -75,10 +75,11 @@ func restamp(
 	return base64.URLEncoding.EncodeToString(payload)
 }

-// decodeCookie feeds value back through the store's decode path.
+// decodeCookie feeds value back through the store's decode path. It
+// asks the store rather than Session.Get, which treats a cookie that
+// does not decode as absent and so hides the codec's reason.
 func decodeCookie(
 	t *testing.T,
-	s *session.Session,
 	value string,
 ) (*sessions.Session, error) {
 	t.Helper()
@@ -94,7 +95,7 @@ func decodeCookie(
 		SameSite: http.SameSiteLaxMode,
 	})

-	sess, err := s.Get(req)
+	sess, err := session.NewStore(testKey()).Get(req, session.SessionName)
 	require.NotNil(t, sess)

 	return sess, err
@@ -105,7 +106,7 @@ func TestCodec_AcceptsCookieInsideAbsoluteCap(t *testing.T) {

 	s := testSession(t)

-	sess, err := decodeCookie(t, s, restamp(
+	sess, err := decodeCookie(t, restamp(
 		t,
 		issuedCookie(t, s),
 		time.Now().Add(-(testAbsoluteMaxAge-time.Hour)),
@@ -126,7 +127,7 @@ func TestCodec_RejectsCookiePastAbsoluteCap(t *testing.T) {

 	s := testSession(t)

-	sess, err := decodeCookie(t, s, restamp(
+	sess, err := decodeCookie(t, restamp(
 		t,
 		issuedCookie(t, s),
 		time.Now().Add(-(testAbsoluteMaxAge+time.Hour)),
@@ -146,8 +146,8 @@ func newStore(key []byte) *sessions.CookieStore {
 //
 // This is decided per-request, not once at startup. Deciding it at
 // startup from the configured environment is what this replaces, and
-// it got the DEFAULT posture wrong: "dev" is the environment when
-// WEBHOOKER_ENVIRONMENT is unset, so a deployment terminating TLS at a
+// it got the DEFAULT posture wrong: "dev" was then the environment when
+// WEBHOOKER_ENVIRONMENT was unset, so a deployment terminating TLS at a
 // proxy without also setting the environment emitted the
 // authentication cookie with no Secure attribute -- silently, and on
 // the same response as a CSRF cookie that did have one.
@@ -224,10 +224,22 @@ func New(
 }

 // Get retrieves a session for the request.
+//
+// A session cookie that does not decode -- one signed with an earlier
+// session key, say, because the database was made anew -- is treated
+// as absent: the caller gets a new, empty session and no error, and
+// the next save replaces the cookie.
 func (s *Session) Get(
 	r *http.Request,
 ) (*sessions.Session, error) {
-	return s.store.Get(r, SessionName)
+	sess, err := s.store.Get(r, SessionName)
+	if sess == nil {
+		return nil, err
+	}
+
+	// For a cookie that does not decode, gorilla/sessions returns a
+	// new, empty session alongside the error that is dropped here.
+	return sess, nil
 }

 // GetKey returns the raw 32-byte authentication key used for
@@ -990,8 +990,8 @@ func sessionCookieFrom(

 // TestSave_SecureFollowsRequestTransport is the regression test for
 // the defect this replaces: Secure was fixed at startup from the
-// configured environment, and "dev" is the environment when
-// WEBHOOKER_ENVIRONMENT is unset. A deployment behind a TLS proxy in
+// configured environment, and "dev" was then the environment when
+// WEBHOOKER_ENVIRONMENT was unset. A deployment behind a TLS proxy in
 // that DEFAULT posture shipped the authentication cookie with no
 // Secure attribute and said nothing about it.
 //
@@ -120,9 +120,12 @@
                        <label class="text-sm text-gray-700">Timeout (seconds, blank = default):</label>
                        <input type="number" name="timeout" min="0" max="300" :disabled="targetType !== 'http'" class="input text-sm w-24">
                    </div>
-                    <div x-show="targetType === 'http'" class="flex gap-2 items-center">
-                        <label class="text-sm text-gray-700">Max retries (0 = fire-and-forget):</label>
-                        <input type="number" name="max_retries" value="0" min="0" max="20" class="input text-sm w-24">
+                    <div x-show="targetType === 'http'">
+                        <div class="flex gap-2 items-center">
+                            <label class="text-sm text-gray-700">Max retries:</label>
+                            <input type="number" name="max_retries" value="0" min="0" max="20" class="input text-sm w-24">
+                        </div>
+                        <p class="text-xs text-gray-500 mt-1">This is the total number of delivery attempts, not retries on top of the first: a value of 3 makes three attempts in all. 0 means a single attempt with no retries and no circuit breaker.</p>
                    </div>
                    <div x-show="targetType === 'slack'">
                        <input type="url" name="url" placeholder="https://hooks.slack.com/services/..." :disabled="targetType !== 'slack'" class="input text-sm">
@@ -69,7 +69,7 @@
            <div class="form-group">
                <label for="max_retries" class="label">Max retries</label>
                <input type="number" id="max_retries" name="max_retries" value="{{.Target.MaxRetries}}" min="0" max="20" class="input">
-                <p class="text-xs text-gray-500 mt-1">0 is fire-and-forget: one attempt, no circuit breaker.</p>
+                <p class="text-xs text-gray-500 mt-1">This is the total number of delivery attempts, not retries on top of the first: a value of 3 makes three attempts in all. 0 means a single attempt with no retries and no circuit breaker.</p>
            </div>
            {{end}}
Author	SHA1	Message	Date
sneak	1647b43aa6	deploy (#365 ) check / check (push) Canceled after 0s Details Reviewed-on: #365	2026-09-29 13:06:27 +02:00
sneakandclawbot	f7151f0168	deployment to prod (#357 ) check / check (push) Successful in 10s Details Co-authored-by: clawbot <35+clawbot@noreply.example.org> Reviewed-on: #357	2026-09-29 12:01:56 +02:00
sneak	3cc05a36eb	deployment (#355 ) check / check (push) Successful in 4m11s Details Reviewed-on: #355	2026-09-29 11:54:06 +02:00
sneak	7316f0a7e2	Merge branch 'main' into prod check / check (push) Successful in 7s Details	2026-09-29 11:49:18 +02:00
clawbotandsneak	5f84d891cf	Deploy: main into prod (#343 ) check / check (push) Successful in 3m32s Details Brings `prod`, which upaas deploys, up to `main` at `9cf9cdd`, the merge of #321. `prod` was cut from `main` at `251cb3d` (1.0.0b1). What it deploys is everything listed in #321. For running it: - With `WEBHOOKER_ENVIRONMENT` unset, the instance runs as `prod` and sends no `Access-Control-Allow-Origin: *`. - Each event database gains its new indexes the first time it is opened after the upgrade. - `webhooker_delivery_retries_total` no longer counts a circuit breaker holding back a delivery that is already `retrying`. Not in this PR yet: #340, in which the container sets its own data directory owner and mode before start. It is in progress on `next`. Once it reaches `main`, this PR carries it, because the PR follows `main`. Model: opus-5-5 Co-authored-by: Jeffrey Paul <1+sneak@noreply.example.org> Reviewed-on: #343	2026-09-29 02:48:03 -07:00
sneak	9cf9cdd8eb	1.0.0 milestone: next into main (#321 ) check / check (push) Successful in 8s Details Reviewed-on: #321	2026-09-29 11:04:58 +02:00
clawbot	f0adeafde3	Drop Set-Cookie from the recovered 500 (closes #193 ) check / check (push) Successful in 3m35s Details When a handler sets a cookie and then panics before sending anything, the recover middleware now deletes Set-Cookie before writing its 500, so a request that failed never hands the client a credential. Every other header, Location included, is left as http.Error leaves it, matching chi's Recoverer. A response that was already sent is untouched. Tests cover the uncommitted case (no cookie, Location kept) and assert the cookie still reaches the client when the response was committed before the panic. Model: opus-5-5	2026-09-29 10:30:26 +02:00
clawbot	f755c03110	Default-block Azure WireServer's public address (closes #245 ) check / check (push) Successful in 4m34s Details Add 168.63.129.16 (Azure WireServer) to blockedNetworks, the default blocklist, not alwaysBlockedNetworks: it is public unicast, so an operator who lists it in ALLOWED_EGRESS_CIDRS can reach it again. The refusal message, the allowlist startup warning, the README and the comments no longer call every blocked address private/reserved, and no longer claim the allowlist cannot open any metadata endpoint. Sources: - https://learn.microsoft.com/en-us/azure/virtual-network/what-is-ip-address-168-63-129-16 - https://learn.microsoft.com/en-us/azure/virtual-machines/metadata-security-protocol/overview Deviation: 147.75.207.243 (Equinix Metal) is not added; Equinix documents only a hostname, and the service was sunset on 2026-06-30. Model: opus-5-5	2026-09-29 10:22:07 +02:00
clawbot	4a724130ca	Close archive writers when the delivery engine stops (closes #280 ) check / check (push) Successful in 3m45s Details The engine cached archive writers and never closed them at shutdown, so after a clean stop an archive's rows could sit in its -wal while the .db held no table. The engine's stop hook now evicts every cached writer once its workers have returned, the same way deleting a webhook does, so a clean stop leaves each archive as one file and a late write is refused. If the workers do not return within the stop budget, the writers are left open as a kill would leave them: closing would wait on a write in progress, and a still-running worker would open new ones. The README no longer says archives keep their sidecars across a clean stop. Model: opus-5-5	2026-09-29 08:30:22 +02:00
clawbot	d4f4ddf51f	Send Content-Type once on a delivery (closes #246 ) check / check (push) Successful in 3m20s Details A delivery set Content-Type from the event's ContentType and then added the inbound Content-Type from the event's stored headers, so a target could receive two values. The inbound Content-Type is no longer forwarded from the stored headers; the receiver already saves it as the event's ContentType. Which value is sent is now stated at applyRequestHeaders: a Content-Type configured on the target, otherwise the event's ContentType, otherwise none. A configured one still survives a cross-origin 307/308 with its body. Model: opus-5-5	2026-09-29 07:11:55 +02:00
clawbot	51580a2bc6	Fail a pending delivery whose target was deleted (closes #293 ) check / check (push) Successful in 3m38s Details Restart recovery and the pending sweep skipped a pending delivery whose target was missing from the batch's target map, every minute, for the life of the database. A miss now asks loadTarget: no row fails the delivery terminally with a recorded reason; any other error leaves it pending, since the map is also empty when its query failed; a target found there is used. The failure goes through the ownership-gated function the retrying paths already used, now failMissingTarget. Once it owns the delivery it re-reads the row and fails it only if the status is unchanged, so a delivery sent and settled in between is left alone. Model: opus-5-5	2026-09-29 06:48:19 +02:00
clawbot	e0b211f960	Make deliveries refused while half-open wait a cooldown (closes #306 ) check / check (push) Successful in 5m0s Details While the breaker was half-open, Allow refused every delivery but the probe and CooldownRemaining returned zero, so each queued task for the target went straight back onto the retry channel and rewrote its status on every pass until the probe finished. CooldownRemaining now returns the whole cooldown while half-open, so a refused delivery waits that long. A refused delivery already at retrying is not written again, so the retry counter now moves only when a refusal moves a delivery into retrying. Model: opus-5-5	2026-09-29 05:48:20 +02:00
clawbot	978eb01b29	Open each event database once when callers race (closes #291 ) check / check (push) Successful in 3m48s Details GetDB opened the database on a cache miss and then tried to cache it, so callers racing on a webhook's first use could each open the file, and the losers closed their copies. On a new file the parallel opens also create its tables at the same time, and one caller can fail with "table already exists". A mutex now covers the open: GetDB looks in the cache again under it, then opens and caches. DeleteDB and CloseAll take the same mutex, so neither runs while an open is under way. Reading an already cached database takes no lock. The new test starts many callers on one webhook at once and checks that exactly one open happened. Model: opus-5-5	2026-09-29 04:37:30 +02:00
clawbot	3cdab97930	Say make check needs make bootstrap on a fresh clone (closes #282 ) check / check (push) Successful in 11s Details The third-party browser assets are not committed, so on a fresh clone make check fails in the tests until make bootstrap (or make assets) has fetched them. The Entrypoints section now says so up front, and why the check does not fetch them itself: it must not change files in the repo. Model: opus-5-5	2026-09-29 04:30:14 +02:00
clawbot	83740b1de1	Do not resend a delivery that restart recovery already sent (closes #299 ) check / check (push) Successful in 3m35s Details Restart recovery could find a just-written delivery pending, send it and release it before the receiver's Notify queued the same delivery. Notify's claim then succeeded on the released id, and the worker sent it again because the new-task path never read the delivery's row. Before sending a new task the worker now reads the delivery's status by primary key and skips the task unless the row still says pending, as the retry path already does for retrying. Nothing else can change the row while the worker owns the delivery. A row left pending by a failed bookkeeping write is still sent again. loadRetryDelivery is renamed loadDelivery now that both paths use it. Model: opus-5-5	2026-09-29 04:12:14 +02:00
sneak	aeeeca5ea1	Merge branch 'main' into next check / check (push) Successful in 9s Details	2026-09-29 03:13:57 +02:00
clawbot	6ebac4fa71	Index the event-tier columns the sweeps, event log and retention scan (closes #314 ) check / check (push) Successful in 3m4s Details The per-webhook event databases had no secondary indexes, so startup recovery, the retry and pending sweeps, the queue-depth sampler, the event log and retention each read whole tables. Indexes declared in the GORM model tags now serve them, and AutoMigrate adds them to new and existing databases alike. Each index also covers deleted_at: GORM adds deleted_at IS NULL to these queries, and SQLite, with no table statistics, otherwise prefers the existing deleted_at index. A test checks SQLite's plan for each statement as GORM builds it. Rule suppressed: lll on the three event-tier model structs, whose struct tags cannot wrap. The resubmitted_from_id scan is left to #325. Model: opus-4-8 (implementation); opus-5-5 (rework)	2026-09-28 14:13:22 +02:00
clawbot	237f131367	Default WEBHOOKER_ENVIRONMENT to prod (closes #307 ) check / check (push) Successful in 3m14s Details An unset WEBHOOKER_ENVIRONMENT now means prod, not dev. The only thing dev still changes is CORS, which then answers every origin with Access-Control-Allow-Origin: *, so an operator who forgets the variable is no longer silently permissive; dev must be set explicitly. Cookie Secure and CSRF strictness follow each request's transport and are unaffected. The README, comments and tests no longer describe dev as the default: the deployment checklist asks only that the environment is not dev, the Docker and nginx examples drop the now-redundant setting, and the TRUSTED_PROXIES warning gives its real reason for firing in every environment. Model: opus-4-8 (implementation); opus-5-5 (rework)	2026-09-28 12:47:31 +02:00
clawbot	7ed1588443	Document running webhooker under upaas (closes #323 ) check / check (push) Successful in 8s Details Adds a "Running under upaas" section to the README: add no port mapping, since upaas publishes mapped ports on every host interface, and put the app on the reverse proxy's Docker network instead; one data volume at /var/lib/webhooker, created owned by UID 1000 before the first deploy; WEBHOOKER_ENVIRONMENT and TRUSTED_PROXIES; the health check upaas reads 60 seconds after a deploy; and where the first-run admin password appears and how to reset it. upaas bind-mounts a host directory it never creates, and one made by root stops the container at its data directory lock. The documented creation step removes that; the image is unchanged. Model: opus-5-5	2026-09-28 12:30:32 +02:00
clawbot	b051821370	Say max_retries is the total attempt count, not a retry count (closes #316 ) check / check (push) Successful in 3m54s Details The help text under the field on both target forms and the max_retries rows in the README now say the number is the total number of delivery attempts: 0 is a single attempt with no retries and no circuit breaker, and N is N attempts in total. The delivery code already worked this way; only the wording was wrong, so an operator wanting one try plus two retries would have entered 2 instead of 3. A UI copy test renders both forms and pins the wording. Delivery behaviour is unchanged. Model: opus-4-8 (implementation); fable-5-1 (merge)	2026-09-21 18:33:09 +02:00
clawbot	39afa69bfc	Consolidate the data directory mode into one owner (closes #288 ) check / check (push) Successful in 6m34s Details Two packages each declared the 0o750 mode for DATA_DIR and both created the directory. internal/datadir now exports DirPerm as the single definition, and internal/database uses it in both places it creates the directory. The value is unchanged, so existing deployments see no permission change. datadir owns it because guarding and creating DATA_DIR is that package's whole purpose and it imports nothing that would form a cycle. Model: opus-4-8 (implementation and review); fable-5-1 (merge)	2026-09-21 10:01:51 +02:00
clawbotandsneak	888eaf526b	State the UUID-is-the-credential rule as a rule (closes #301 ) (#302 ) check / check (push) Successful in 4m28s Details Closes #301. Docs-only apart from one test comment; no behaviour change. The receiver has authenticated on the entrypoint UUID alone since inbound signature verification was removed in #279. The README described that as the current state. It did not say it is the decision, which leaves a future contributor free to propose HMAC as an improvement rather than as a reversal. What changed: - `## The entrypoint URL is the authentication secret` now states the rule: the v4 UUID at `/webhook/{uuid}` is the credential and the only one; no shared secret, HMAC signature, bearer token or second factor will be added, including as defence in depth. It names the removal that settled it, and it says explicitly that signature headers a sender sends anyway are stored and forwarded but never checked — the previous text left that ambiguous. - The same section carries the two consequences an operator has to act on: the URL is a capability, so keep it out of logs, tickets and screenshots; and rotation means minting a new entrypoint, not changing a key. - It also handles the case the rule will next be argued from: a sender that only supports signed payloads to a well-known URL is a constraint on that integration, to be raised on its own terms, not grounds to reintroduce shared secrets. - The rule is reachable without scrolling 1,100 lines: a pointer in the intro, a new first bullet under Authentication (which previously listed the web UI, the API and `/metrics` and said nothing about the receiver at all), and a sharpened bullet under Security. Stale language found and corrected: one, in `internal/delivery/redirect_test.go`. Its comment justified same-origin header retention partly by "the inbound signature the receiver verifies" — in this repo's vocabulary "the receiver" is `/webhook/{uuid}`, which verifies nothing. The endpoint that verifies it is the delivery target's, and the comment now says so. Two places that read like stale signing language were checked and left alone as accurate: `internal/delivery/redirect.go` and `internal/server/sentry.go` describe signature headers senders put on the receiver route, which do arrive and are forwarded — neither claims webhooker checks them. `REPO_POLICIES.md` was deliberately not touched. It is the cross-project policy document synced from `sneak/prompts` and carries `last_modified` front matter for that purpose, so a webhooker-specific carve-out does not belong in it. Worth knowing: its hardening section ends "if a standard security hardening measure exists for HTTP services and is not listed here, it is still expected. When in doubt, harden" — that is the sentence a future HMAC proposal will cite, and only the README now answers it. `TODO.md` is untouched per its own Workflow section (issue branches do not touch it). Co-authored-by: sneak <sneak@sneak.berlin> Reviewed-on: #302 Co-authored-by: clawbot <clawbot@noreply.example.org> Co-committed-by: clawbot <clawbot@noreply.example.org>	2026-08-30 04:05:38 +02:00
sneak	251cb3d3d3	1.0.0b1 - testing remains check / check (push) Successful in 8s Details Reviewed-on: #111	2026-08-30 04:04:59 +02:00
clawbot	d61d9dc1c1	Drop the stale open-work claim from the TODO status section check / check (push) Successful in 2m56s Details The milestone is the authoritative count and the section already says so; the lead-in asserted work remaining independently of it.	2026-08-24 15:52:44 +00:00
clawbot	b0a011f6b4	Render unknown for a zero CreatedAt in Slack/Mattermost messages (closes #298 ) check / check (push) Superseded by a newer commit; never tested	2026-08-24 17:52:25 +02:00
clawbot	5976a4a98f	Carry the event's receipt time into every delivery (closes #257 ) check / check (push) Successful in 3m4s Details	2026-08-24 06:44:24 +02:00
clawbot	b2c9acdaa6	Correct four documentation claims ahead of the 1.0.0 tag check / check (push) Successful in 7s Details	2026-08-24 06:25:08 +02:00
clawbot	af3703d748	Close the two remaining delivery terminal-state gaps (closes #107 ) check / check (push) Successful in 3m16s Details	2026-08-24 05:12:02 +02:00
clawbot	322d9a6d6b	Create every SQLite file 0600 (closes #255 ) check / check (push) Successful in 3m17s Details	2026-08-24 04:33:40 +02:00
clawbot	b9f7db6901	Fail loudly on an unparseable SENTRY_DSN and a malformed .env (closes #283 ) check / check (push) Successful in 2m54s Details	2026-08-24 04:25:29 +02:00
clawbot	48cf93ec7e	Derive the entrypoint URL scheme from the shared TLS predicate (closes #272 ) check / check (push) Successful in 3m44s Details	2026-08-24 04:04:04 +02:00
clawbot	8d64259283	Open SQLite with WAL and bound delivery re-dispatch by ownership (closes #256 ) check / check (push) Successful in 3m1s Details	2026-08-24 03:49:17 +02:00
clawbot	bde32d3ee6	Record landed 1.0.0 work and drop the removed signing roadmap in TODO.md check / check (push) Successful in 8s Details	2026-08-24 03:43:16 +02:00
clawbot	62576f6fc6	Bind the app port deliberately and document the proxy deployment (closes #268 ) (closes #226 ) check / check (push) Successful in 3m4s Details	2026-08-24 03:38:44 +02:00