Point to the proxy's access log for a login flood's source

No log line records the client address taken from X-Forwarded-For,
so the login endpoint section no longer says the source shows in the
failure logs; it names the proxy's access log instead.

Model: opus-5-5

README.md

@@ -147,7 +147,7 @@ TTY detection, and security headers are always applied.
 | `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` |
-| `TRUSTED_PROXIES`       | CIDRs whose forwarded headers are trusted (unset: all clients behind a proxy share one rate-limit bucket; a correct login password is never throttled either way) | `""` (none) |
+| `TRUSTED_PROXIES`       | CIDRs whose forwarded headers are trusted. A set value replaces the default. Under the default, any client with a private address, whether it connects directly or through the proxy, can choose its own rate-limit key by sending its own `X-Forwarded-For`; if any clients have private addresses, set it to the proxy's address alone. See [Trusted proxies](#trusted-proxies) | `10.0.0.0/8,172.16.0.0/12,192.168.0.0/16` (RFC 1918) |
 | `ALLOWED_EGRESS_CIDRS`  | CIDRs that delivery targets may reach despite the SSRF blocklist. Read [Allowing egress to your own network](#allowing-egress-to-your-own-network) before setting it | `""` (none) |
 
 #### Allowing egress to your own network
@@ -157,6 +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
@@ -195,15 +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 |
   | ----------------------- | ---------- |
@@ -242,7 +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.
@@ -372,41 +379,48 @@ unlocked.
 `TRUSTED_PROXIES` is a comma-separated list of CIDR blocks (a bare
 address such as `192.168.1.7` is accepted and treated as a single
 host), for example `192.168.1.7, 2001:db8::5`. It decides whose
-`X-Forwarded-For` header the rate limiters believe, so it should name
-the addresses of your reverse proxies and nothing else.
+`X-Forwarded-For` header the rate limiters believe, so it should cover
+the addresses of your reverse proxies.
 
 `X-Forwarded-For` is honoured **only** when the connecting peer is
 inside one of these blocks; for every other peer the client identity is
-the connection's own address and the header is ignored. The default is
-the empty list, which trusts nobody — anything else would let any
-client pick its own rate limit bucket, minting a fresh one per request
-or draining someone else's. Set it to the address of your reverse
-proxy, and to nothing wider. A set but unparseable value aborts
-startup.
+the connection's own address and the header is ignored. Unset (or
+empty), the list is the RFC 1918 private ranges: `10.0.0.0/8`,
+`172.16.0.0/12` and `192.168.0.0/16`. That covers a reverse proxy
+reaching webhooker over a Docker network or a private LAN without
+anything set. A set value replaces the default entirely. A set but
+unparseable value aborts startup.
 
-That default is safe against forged headers, but leaving it unset in
-production has a cost you must know about. Production runs behind a
-TLS-terminating reverse proxy, so with `TRUSTED_PROXIES` unset every
-request keys on the proxy's own address and all clients share a single
+Trusting those ranges has two consequences for clients with private
+addresses:
+
+- Any such client, whether it connects directly or through the proxy,
+  can choose its own rate-limit key by sending its own
+  `X-Forwarded-For`. A direct client's header is walked because the
+  client is itself trusted; behind the proxy, the client's own address
+  is skipped as a trusted hop when the chain is walked (below), so the
+  entry it wrote is taken as the client. If any of your clients have
+  private addresses, you must set `TRUSTED_PROXIES` to the proxy's
+  address alone.
+- A client behind the proxy that sends no `X-Forwarded-For` of its own
+  shares the proxy's bucket, because its own address is skipped too.
+  Setting the list to the proxy's address alone gives each its own
+  bucket.
+
+A proxy the list does not cover, such as nginx on the same host
+reaching webhooker over loopback, is not trusted: every request through
+it keys on the proxy's own address and all clients share a single
 bucket per limit. The receiver limits become service-wide ceilings,
 and the login endpoint's failure counting collapses onto one key, so a
 stranger's wrong passwords throttle every other client's wrong
-passwords.
+passwords. Set `TRUSTED_PROXIES` to that proxy's address to restore
+per-client buckets.
 
 What it cannot do is lock the operator out. The login endpoint
 verifies credentials **before** it consults any limit and charges only
 failures, so a correct password is never throttled no matter how full
 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,
-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.
-
 `X-Real-IP` and `True-Client-IP` are **never** read, from any peer.
 Reverse proxies append to `X-Forwarded-For` but forward other client
 headers verbatim, so a single-valued header is client-controlled even
@@ -428,14 +442,14 @@ Two operator requirements follow:
   (nginx `$proxy_add_x_forwarded_for`, HAProxy `option forwardfor`,
   Caddy and AWS ALB by default), and must append a bare address with
   no port.
-- List proxy hosts **only**. Any address inside `TRUSTED_PROXIES`
+- Keep clients out of the list. Any address inside `TRUSTED_PROXIES`
   chooses its own rate-limit key: its `X-Forwarded-For` is walked, so
   it can name a different address on every request to get a fresh
   bucket each time, or name another client's address to drain that
-  client's bucket. Never list a block that also covers clients — a
-  broad `10.0.0.0/8` on a network where clients live in the same range
-  makes all three limits, including the unauthenticated webhook
-  receiver, silently bypassable by every client in the block.
+  client's bucket. A block that also covers clients — the default, on
+  a network where clients have private addresses — makes all three
+  limits, including the unauthenticated webhook receiver, silently
+  bypassable by every client in the block.
 
 #### Sessions
 
@@ -757,10 +771,16 @@ repository's `Dockerfile` and runs it. The app needs:
 
 - **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).
+    - `TRUSTED_PROXIES`: Docker networks use private addresses, so the
+      default covers your reverse proxy on that network. Under the
+      default, any client with a private address, whether it connects
+      directly or through the proxy, can choose its own rate-limit key
+      by sending its own `X-Forwarded-For`. If any clients have private
+      addresses, or the network's addresses are outside the RFC 1918
+      ranges, set it to the proxy's address there. 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`.
@@ -823,12 +843,14 @@ reports.
    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.
+3. **Make sure `TRUSTED_PROXIES` covers the proxy's address.** Unset,
+   it covers the RFC 1918 private ranges, so a proxy on a Docker
+   network or a private LAN is covered and one on loopback is not. For
+   a proxy it does not cover, every rate limiter keys on the connecting
+   peer, which 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). If
+   any clients have private addresses, 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
@@ -968,15 +990,10 @@ scratch file**: it holds committed transactions that are not yet in the
 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 `webhooker.db` and every `events-*.db`, which
-checkpoints and removes their sidecars; a killed or crashed instance
-leaves them, and they must be carried with the `.db`. **Archive
-databases are different**: their handle is not closed at shutdown, so
-`archive-*.db-wal` and `-shm` normally survive a clean stop and the
-`-wal` can hold every row the archive has. Measured on a stopped
-instance: `archive-….db` 4096 bytes with no table, its `-wal` 157 KB
-holding all 8 archived events. Copying `DATA_DIR` in full is what makes
-this a non-issue; copying `.db` files out of it by name is not.
+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
@@ -1051,10 +1068,9 @@ 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.
-Shutdown is **not** on that list: the archive handle is not closed when
-the service stops. So either move `archive-{uuid}.db` together with any
-`-wal`/`-shm` beside it, or wait until there are none.
+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
 
@@ -1073,12 +1089,10 @@ the service stops. So either move `archive-{uuid}.db` together with any
    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 has none for `webhooker.db` or the `events-*.db`,
-   because a clean stop closes those and checkpoints their sidecars
-   away — but it will normally have them for `archive-*.db`, whose
-   handle stays open across shutdown, and those carry the archive's
-   rows. A copy salvaged from a crashed instance has them for
-   everything, and needs all of them.
+   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
@@ -1397,10 +1411,11 @@ It uses:
 - **[go-chi/httprate](https://github.com/go-chi/httprate)** for
   sliding-window rate limiting of the password-change and webhook
   receiver endpoints. The bucket is per client IP only when
-  `TRUSTED_PROXIES` names the reverse proxy; unset, every client
-  behind that proxy shares one bucket per limit. The login endpoint
-  counts failed attempts itself instead, so that a correct password is
-  never throttled (see [Rate Limiting](#rate-limiting))
+  `TRUSTED_PROXIES` covers the reverse proxy (by default it covers the
+  RFC 1918 private ranges); otherwise every client behind that proxy
+  shares one bucket per limit. The login endpoint counts failed
+  attempts itself instead, so that a correct password is never
+  throttled (see [Rate Limiting](#rate-limiting))
 - **[Prometheus](https://prometheus.io)** for metrics, served at
   `/metrics` behind basic auth
 - **[Sentry](https://sentry.io)** for optional error reporting
@@ -2581,30 +2596,30 @@ let one subscriber rotate source addresses and mint a fresh bucket per
 request, evading these limits at the network layer without spoofing
 anything; the cost is that distinct clients inside one `/64` share a
 bucket. IPv4-mapped addresses (`::ffff:1.2.3.4`) key as the IPv4 address
-they carry. See [Trusted proxies](#trusted-proxies). Deployed without that
-variable set, a client behind a reverse proxy shares one bucket with
-every other client behind the same proxy. Set `TRUSTED_PROXIES` to the
-proxy's address to get per-client limits back. What the shared bucket
+they carry. See [Trusted proxies](#trusted-proxies). When that variable
+does not cover the reverse proxy, a client behind it shares one bucket
+with every other client behind the same proxy. Set `TRUSTED_PROXIES` to
+the proxy's address to get per-client limits back. What the shared bucket
 costs is not the same for every limiter, and the two cases pull in
 opposite directions:
 
 - For the **receiver** limits it costs throughput, which is the safe
   direction to be wrong in: sharing can only make a limit bind sooner,
   never let a sender past it. It matters more for the aggregate limit
-  than for the per-entrypoint one: with `TRUSTED_PROXIES` unset behind
-  the reverse proxy a production deployment is required to run behind,
-  every request keys on the proxy, so the aggregate limit becomes a
-  service-wide ceiling of 1200 requests per minute across all senders
-  and all entrypoints, where the per-entrypoint limit's capacity still
-  grows with the number of entrypoints. Any deployment with more than a
-  handful of busy entrypoints must set `TRUSTED_PROXIES`.
+  than for the per-entrypoint one: when `TRUSTED_PROXIES` does not
+  cover the reverse proxy a production deployment is required to run
+  behind, every request keys on the proxy, so the aggregate limit
+  becomes a service-wide ceiling of 1200 requests per minute across all
+  senders and all entrypoints, where the per-entrypoint limit's
+  capacity still grows with the number of entrypoints. Any deployment
+  with more than a handful of busy entrypoints must make sure
+  `TRUSTED_PROXIES` covers its proxy.
 - For the **login and password-change** limits it costs precision, not
   availability. Login failures from every client land in one counter,
   so a stranger's wrong passwords make the operator's own wrong
   passwords answer `429` sooner; the operator's _correct_ password is
   never affected, because it is never counted. Production deployments
-  should still set `TRUSTED_PROXIES`; webhooker warns at startup
-  whenever it is empty, in any environment.
+  should still make sure `TRUSTED_PROXIES` covers their proxy.
 
 #### The login endpoint
 
@@ -2707,8 +2722,10 @@ re-fills both verification slots on its first two requests. The
 remedies are to block the source at the reverse proxy, or to
 rate-limit `POST /pages/login` there — the one place a limit can be
 applied without reintroducing the lockout, because the proxy sees the
-real client address. Setting `TRUSTED_PROXIES` does not stop the
-saturation, but it makes the source visible in the failure logs.
+real client address. `TRUSTED_PROXIES` does not stop the saturation.
+The flood's source is in the proxy's access log: webhooker's own logs
+record the proxy's address, not the client's (see
+[Deployment behind a reverse proxy](#deployment-behind-a-reverse-proxy)).
 
 Finer-grained per-webhook rate limits (configured in the web UI and
 enforced in the webhook handler) can layer on top of this env-level
@@ -3065,10 +3082,9 @@ check, see [The login endpoint](#the-login-endpoint).
   It runs behind session auth, so only a client already holding a
   valid session reaches it, and an operator throttled out of changing
   a password can still log in. The bucket is per client IP only when
-  `TRUSTED_PROXIES` names the reverse proxy; unset, every client
+  `TRUSTED_PROXIES` covers the reverse proxy; otherwise every client
   shares one bucket, which costs precision rather than availability
-  (see [Rate Limiting](#rate-limiting)). webhooker warns at startup
-  whenever `TRUSTED_PROXIES` is empty
+  (see [Rate Limiting](#rate-limiting))
 - Prometheus metrics behind basic auth
 - Static assets embedded in binary (no filesystem access needed at
   runtime)
@@ -3088,7 +3104,8 @@ 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

internal/config/config.go

@@ -75,6 +75,11 @@ const (
 	// internet-exposed endpoint.
 	defaultReceiverRateLimit = 120
 
+	// defaultTrustedProxies is TRUSTED_PROXIES when it is unset: the
+	// RFC 1918 private ranges, which a reverse proxy reaching the
+	// process over a Docker network or a private LAN connects from.
+	defaultTrustedProxies = "10.0.0.0/8,172.16.0.0/12,192.168.0.0/16"
+
 	// maxPort is the highest valid TCP port number. The lower
 	// bound (at least 1) is enforced by envPositiveInt.
 	maxPort = 65535
@@ -172,13 +177,14 @@ type Config struct {
 
 	// TrustedProxies is the set of networks whose members are
 	// allowed to speak for the client with X-Forwarded-For, the
-	// only forwarded header read. It is empty unless
-	// TRUSTED_PROXIES is set, and empty means no peer is
-	// trusted: forwarded headers are then ignored entirely and
-	// clients are identified by the connection's own address.
-	// Members can choose their own rate-limit key, so this must
-	// name proxy hosts only, never a block that also covers
-	// clients.
+	// only forwarded header read. Unless TRUSTED_PROXIES is set it
+	// is the RFC 1918 private ranges (defaultTrustedProxies).
+	// Other peers' forwarded headers are ignored and they are
+	// identified by the connection's own address. Under the
+	// default any client with a private address, directly or
+	// through a proxy, can choose its own rate-limit key, so
+	// where any clients have private addresses this must be set
+	// to the proxy hosts alone.
 	TrustedProxies []netip.Prefix
 
 	// AllowedEgressCIDRs is the set of networks a delivery target
@@ -192,9 +198,10 @@ 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
@@ -459,14 +466,15 @@ func parseCIDR(entry string) (netip.Prefix, error) {
 
 // envPrefixList returns the value of the named environment variable
 // parsed as a comma-separated list of CIDR blocks (bare addresses
-// allowed). An unset, empty, or blank value yields an empty list. A
-// set value containing an unparseable entry is a hard error naming
-// the key and the bad entry, so startup fails loudly rather than
-// silently running with a list the operator did not intend.
-func envPrefixList(key string) ([]netip.Prefix, error) {
+// allowed). An unset, empty, or blank value is read as defaultValue
+// instead. A set value containing an unparseable entry is a hard
+// error naming the key and the bad entry, so startup fails loudly
+// rather than silently running with a list the operator did not
+// intend.
+func envPrefixList(key, defaultValue string) ([]netip.Prefix, error) {
 	v := strings.TrimSpace(os.Getenv(key))
 	if v == "" {
-		return nil, nil
+		v = defaultValue
 	}
 
 	var prefixes []netip.Prefix
@@ -680,12 +688,12 @@ func loadFromEnv() (*Config, error) {
 		return nil, err
 	}
 
-	trustedProxies, err := envPrefixList("TRUSTED_PROXIES")
+	trustedProxies, err := envPrefixList("TRUSTED_PROXIES", defaultTrustedProxies)
 	if err != nil {
 		return nil, err
 	}
 
-	allowedEgressCIDRs, err := envPrefixList("ALLOWED_EGRESS_CIDRS")
+	allowedEgressCIDRs, err := envPrefixList("ALLOWED_EGRESS_CIDRS", "")
 	if err != nil {
 		return nil, err
 	}
@@ -746,61 +754,19 @@ 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), ","),
 	)
 }
 
-// warnSharedRateLimitBucket logs a startup warning whenever
-// TRUSTED_PROXIES is empty, in any environment.
-//
-// With no trusted proxies every rate limiter keys on the connecting
-// peer's address. Whether that is harmless or dangerous depends on
-// what is in front of the process, which this code cannot observe:
-// with nothing in front, the peer is the client and the limits are
-// per-client as intended; behind a reverse proxy the peer is the proxy
-// for every request, so all clients share one bucket per limiter.
-//
-// The login endpoint no longer spends budget on arrival — it verifies
-// credentials first and charges only failures — so a shared bucket
-// cannot deny the operator a correct password. What it does collapse
-// is the failure counting: one client's wrong passwords throttle
-// everyone else's wrong passwords, and the receiver's limits become
-// service-wide ceilings.
-//
-// 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 —
-// so this warns rather than failing startup or changing the key.
-func (c *Config) warnSharedRateLimitBucket(log *slog.Logger) {
-	if len(c.TrustedProxies) > 0 {
-		return
-	}
-
-	log.Warn(
-		"TRUSTED_PROXIES is empty: every rate limit keys on the "+
-			"connecting peer's address. With nothing proxying to "+
-			"this process that is the client itself and the limits "+
-			"are per-client as intended. Behind a reverse proxy the "+
-			"peer is the proxy on every request, so all clients "+
-			"share one bucket per limit: the receiver limits become "+
-			"service-wide ceilings, and one client's failed logins "+
-			"throttle every other client's failed logins — a "+
-			"correct password still gets in. If anything proxies to "+
-			"this process, set TRUSTED_PROXIES to its address.",
-		"environment", c.Environment,
-		"trustedProxies", len(c.TrustedProxies),
-	)
-}
-
 // New creates a Config by reading environment variables.
 //
 //nolint:revive // lc parameter is required by fx even if unused.
@@ -846,7 +812,6 @@ func New(lc fx.Lifecycle, params ConfigParams) (*Config, error) {
 		"hasMetricsAuth", s.MetricsAuthEnabled(),
 	)
 
-	s.warnSharedRateLimitBucket(log)
 	s.warnEgressAllowlist(log)
 
 	return s, nil

internal/config/config_test.go

@@ -551,6 +551,11 @@ func testReceiverRateLimitSuccess(
 }
 
 func TestTrustedProxies(t *testing.T) {
+	// Unset, the RFC 1918 private ranges are trusted, so a reverse
+	// proxy on a Docker network or a private LAN is covered without
+	// configuration.
+	defaultProxies := []string{cidrPrivateV4, "172.16.0.0/12", "192.168.0.0/16"}
+
 	tests := []struct {
 		name        string
 		set         bool
@@ -559,18 +564,21 @@ func TestTrustedProxies(t *testing.T) {
 		expected    []string
 	}{
 		{
-			// The default must be "trust nobody": an empty list
-			// means forwarded headers are ignored, never that
-			// every peer may speak for the client.
 			name:     caseUnsetUsesDefault,
 			set:      false,
-			expected: []string{},
+			expected: defaultProxies,
 		},
 		{
-			name:     "blank value trusts nothing",
+			name:     "blank value uses default",
 			set:      true,
 			value:    "  ",
-			expected: []string{},
+			expected: defaultProxies,
+		},
+		{
+			name:     "set value replaces the default entirely",
+			set:      true,
+			value:    "203.0.113.7",
+			expected: []string{"203.0.113.7/32"},
 		},
 		{
 			name:  caseValidValueParsed,
@@ -834,107 +842,13 @@ 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")
-		})
-	}
-}
-
-// TestSharedRateLimitBucketWarning covers the startup warning that
-// 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, 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
-		environment    string
-		trustedProxies string
-		expectWarning  bool
-	}{
-		{
-			name:          "prod without trusted proxies warns",
-			environment:   config.EnvironmentProd,
-			expectWarning: true,
-		},
-		{
-			name:           "prod with trusted proxies is quiet",
-			environment:    config.EnvironmentProd,
-			trustedProxies: cidrPrivateV4,
-			expectWarning:  false,
-		},
-		{
-			name:          "dev without trusted proxies warns",
-			environment:   config.EnvironmentDev,
-			expectWarning: true,
-		},
-		{
-			name:           "dev with trusted proxies is quiet",
-			environment:    config.EnvironmentDev,
-			trustedProxies: cidrPrivateV4,
-			expectWarning:  false,
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			// Cannot use t.Parallel() here because t.Setenv
-			// is incompatible with parallel subtests.
-			t.Setenv("WEBHOOKER_ENVIRONMENT", tt.environment)
-
-			if tt.trustedProxies == "" {
-				require.NoError(
-					t, os.Unsetenv("TRUSTED_PROXIES"),
-				)
-			} else {
-				t.Setenv("TRUSTED_PROXIES", tt.trustedProxies)
-			}
-
-			var buf bytes.Buffer
-
-			log := slog.New(slog.NewJSONHandler(
-				&buf, &slog.HandlerOptions{
-					Level: slog.LevelDebug,
-				},
-			))
-
-			require.NoError(
-				t,
-				config.WarnSharedRateLimitBucketForTest(log),
-			)
-
-			if !tt.expectWarning {
-				assert.Empty(t, buf.String())
-
-				return
-			}
-
-			logged := buf.String()
-
-			assert.Contains(t, logged, `"level":"WARN"`)
-			assert.Contains(t, logged, "TRUSTED_PROXIES")
-			assert.Contains(t, logged, "share one bucket")
-			assert.Contains(
-				t, logged, "throttle every other client's failed logins",
-			)
-			// The warning must not claim a lockout the login
-			// endpoint no longer permits: credentials are verified
-			// before any budget is spent.
-			assert.Contains(
-				t, logged, "a correct password still gets in",
-			)
-			// The text must stay accurate for a developer with
-			// nothing in front of the process, where an empty
-			// list costs nothing.
-			assert.Contains(
-				t, logged, "nothing proxying to this process",
-			)
+			// 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")
 		})
 	}
 }

internal/config/export_test.go

@@ -6,21 +6,6 @@ import "log/slog"
 // the external config_test package so each helper can be covered by
 // its own table-driven test without weakening the package API.
 
-// WarnSharedRateLimitBucketForTest loads a Config from the current
-// environment and emits its startup warnings to log. The real logger
-// writes to stdout, so this lets the warning's firing condition be
-// asserted against a handler the test controls.
-func WarnSharedRateLimitBucketForTest(log *slog.Logger) error {
-	c, err := loadFromEnv()
-	if err != nil {
-		return err
-	}
-
-	c.warnSharedRateLimitBucket(log)
-
-	return nil
-}
-
 // WarnEgressAllowlistForTest loads a Config from the current
 // environment and emits its egress-allowlist startup warning to
 // log, so a test can assert both that the warning fires only when

internal/delivery/engine.go

@@ -362,6 +362,15 @@ func (e *Engine) start() {
 // stop cancels the worker pool's context and waits for the pool
 // to drain, bounded by the stop hook's context: a wedged worker
 // must not hang the process past fx's stop timeout.
+//
+// Once the pool has drained it closes the archive writers, so a
+// clean stop leaves no archive -wal behind. Nothing else holds a
+// writer for long by then: the archive sweeper stops before the
+// engine, and deleting a webhook only closes one. If the pool did
+// not drain in time, the writers are left open, as a kill would
+// leave them. Closing them would wait for any write in progress,
+// and a worker still running would then open new writers that
+// nothing closes, so it gains nothing over a kill.
 func (e *Engine) stop(ctx context.Context) error {
 	e.log.Info("delivery engine stopping")
 
@@ -376,6 +385,8 @@ func (e *Engine) stop(ctx context.Context) error {
 		return err
 	}
 
+	e.dbTarget.evictAll()
+
 	e.log.Info("delivery engine stopped")
 
 	return nil

internal/delivery/engine_lifecycle_test.go

@@ -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",
+	)
+}

internal/delivery/ssrf.go

@@ -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.

internal/delivery/ssrf_allowlist_test.go

@@ -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

internal/delivery/target_database.go

@@ -277,6 +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

internal/delivery/target_database_evict_test.go

@@ -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",
+	)
+}

internal/handlers/auth.go

@@ -103,9 +103,10 @@ func (h *Handlers) renderLoginError(
 // The credential check runs BEFORE any rate-limit budget is
 // consulted, and only a failed check spends budget. That is what
 // keeps the single administrative path reachable: behind the reverse
-// proxy this deployment requires, with TRUSTED_PROXIES unset, every
-// client shares one bucket, so a limiter spent on arrival lets any
-// stranger deny the operator's own correct password indefinitely.
+// proxy this deployment requires, when TRUSTED_PROXIES does not cover
+// it, every client shares one bucket, so a limiter spent on arrival
+// lets any stranger deny the operator's own correct password
+// indefinitely.
 //
 // Verifying first means every login POST costs an Argon2id hash, so
 // the work is taken under a bounded number of verification slots.

internal/handlers/auth_test.go

@@ -25,7 +25,7 @@ const (
 
 	// sharedProxyPeer is the whole point of this file. Production is
 	// required to run behind a TLS-terminating reverse proxy, and
-	// TRUSTED_PROXIES defaults to empty, so every client — attacker
+	// when TRUSTED_PROXIES does not cover it every client — attacker
 	// and operator alike — reaches the process from the proxy's
 	// address and shares one rate-limit bucket. Both parties in
 	// these tests therefore use the same RemoteAddr.
@@ -115,11 +115,11 @@ func floodFailures(
 // done-criterion of https://git.eeqj.de/sneak/webhooker/issues/150.
 //
 // The attacker and the operator share one rate-limit bucket, because
-// behind the mandated reverse proxy with TRUSTED_PROXIES unset every
-// client keys on the proxy's address. The attacker floods the
-// operator's own username — a single-admin product has a predictable
-// one — far past the failure limit. The operator must still be able
-// to log in with the correct password.
+// behind the mandated reverse proxy, when TRUSTED_PROXIES does not
+// cover it, every client keys on the proxy's address. The attacker
+// floods the operator's own username — a single-admin product has a
+// predictable one — far past the failure limit. The operator must
+// still be able to log in with the correct password.
 //
 // This fails if credentials stop being verified ahead of the limiter.
 func TestLogin_StrangersFloodCannotLockOutTheOperator(t *testing.T) {

internal/middleware/loginguard.go

@@ -108,10 +108,10 @@ type failureWindow struct {
 //
 // A limiter that spends budget on arrival cannot protect a
 // single-admin product: behind the reverse proxy the deployment
-// requires, with TRUSTED_PROXIES unset, every client keys on the
-// proxy, so a stranger trickling five POSTs a minute keeps the one
-// bucket full and the operator's own correct password is answered 429
-// forever. There is no second administrative path.
+// requires, when TRUSTED_PROXIES does not cover it, every client
+// keys on the proxy, so a stranger trickling five POSTs a minute
+// keeps the one bucket full and the operator's own correct password
+// is answered 429 forever. There is no second administrative path.
 //
 // So budget is spent only by a FAILED verification. A correct
 // password is never throttled, whatever the counters say, which is

internal/middleware/ratelimit.go

@@ -123,9 +123,8 @@ func bucketKey(addr netip.Addr) string {
 	return prefix.String()
 }
 
-// isTrustedProxy reports whether addr belongs to a network the
-// operator listed in TRUSTED_PROXIES. The list is empty by default,
-// so by default nothing is trusted.
+// isTrustedProxy reports whether addr belongs to a network in
+// TRUSTED_PROXIES, which by default is the RFC 1918 private ranges.
 func (m *Middleware) isTrustedProxy(addr netip.Addr) bool {
 	for _, prefix := range m.params.Config.TrustedProxies {
 		if prefix.Contains(addr) {

internal/middleware/ratelimit_test.go

@@ -426,8 +426,8 @@ func assertSharedBucket(
 }
 
 // TestRateLimitKey_SpoofedForwardedFromUntrustedPeer is the test
-// this gating exists for: with no trusted proxies configured (the
-// default), a client that rotates a forwarded header on every
+// this gating exists for: from a peer that is not a trusted
+// proxy, a client that rotates a forwarded header on every
 // request must stay in one bucket. If forwarded headers were
 // trusted unconditionally, each spoofed value would mint a fresh
 // bucket and the limit would stop no one.

internal/middleware/recoverer.go

@@ -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(

internal/middleware/recoverer_test.go

@@ -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"])

internal/server/routes.go

@@ -140,11 +140,12 @@ func (s *Server) setupPageRoutes() {
 		r.Use(s.mw.NoCache())
 
 		// The login POST carries no pre-emptive rate limiter. Behind
-		// the reverse proxy production requires, with TRUSTED_PROXIES
-		// unset, every client shares one bucket, so a limiter spent
-		// on arrival lets any stranger deny the operator the only
-		// administrative path. The handler verifies credentials first
-		// and charges only failures; see Handlers.authenticateUser.
+		// the reverse proxy production requires, when TRUSTED_PROXIES
+		// does not cover it, every client shares one bucket, so a
+		// limiter spent on arrival lets any stranger deny the operator
+		// the only administrative path. The handler verifies
+		// credentials first and charges only failures; see
+		// Handlers.authenticateUser.
 		r.Get("/login", s.h.HandleLoginPage())
 		r.Post("/login", s.h.HandleLoginSubmit())