resolver: a nameserver with a failed record type is not nodata (closes #253)

classifyResponse set nodata when every record type that answered had
no records, even when another type's query got no usable reply. That
type is listed in FailedTypes and its records are unknown, so the
nameserver has not said it has none. It now stays ok, the status
README describes for a nameserver with a failed type, and nodata is
set only when no type failed. The watcher saved nodata as ok already,
so saved state is unchanged; the live test that rejects nodata no
longer fails when one of a nameserver's queries is lost.

Model: opus-5-5

.dockerignore

@@ -1,6 +1,9 @@
-.git/
+# .git is sent, without its config: the builder stage derives the version it
+# stamps into the binary from it, and `git describe` does not need the config,
+# which can hold a credential (a password in the remote URL, a CI token). No
+# tracked file may be listed here: git in the build would see it as deleted
+# and mark the version -dirty, and an excluded .md would silently drop out of
+# the prettier check in Dockerfile.fmt.
+.git/config
 bin/
-*.md
+node_modules/
-LICENSE
-.editorconfig
-.gitignore

.gitea/workflows/check.yml

@@ -1,9 +1,17 @@
 name: check
 on: [push]
+# A new push to a branch cancels that branch's older run, queued or running;
+# runs on other branches, `next` and `main` among them, are left alone.
+concurrency:
+    group: ${{ github.workflow }}-${{ github.ref }}
+    cancel-in-progress: true
 jobs:
     check:
         runs-on: ubuntu-latest
         steps:
             # actions/checkout v4.2.2, 2026-02-28
             - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+              # script/cibuild needs no token, so none is left in .git/config.
+              with:
+                  persist-credentials: false
             - run: script/cibuild

.gitignore

@@ -1,4 +1,5 @@
 bin/
+node_modules/
 vendor/
 data/
 .env

.prettierignore

@@ -0,0 +1,5 @@
+bin/
+data/
+node_modules/
+.claude/
+static/css/tailwind.min.css

.prettierrc

@@ -0,0 +1,4 @@
+{
+    "tabWidth": 4,
+    "proseWrap": "always"
+}

Dockerfile

@@ -1,7 +1,9 @@
 # Lint stage - fast feedback on lint issues, before the build starts.
 # The linter is invoked directly rather than through `make lint`: that
 # target shells out to `docker build -f Dockerfile.lint`, and there is
-# no docker daemon inside a docker build.
+# no docker daemon inside a docker build. For the same reason this stage
+# runs only the Go half of `make fmt-check`; script/cibuild runs the
+# markdown half after this build.
 # script/cibuild and script/docker name this stage in --no-cache-filter.
 # golangci/golangci-lint:v2.12.2 (Debian-based), 2026-08-10
 FROM golangci/golangci-lint:v2.12.2@sha256:5cceeef04e53efe1470638d4b4b4f5ceefd574955ab3941b2d9a68a8c9ad5240 AS lint
@@ -12,7 +14,7 @@ RUN go mod download
 
 COPY . .
 
-RUN make fmt-check
+RUN script/fmt-check-go
 RUN golangci-lint run --config .golangci.yml ./...
 
 # Build stage
@@ -22,6 +24,11 @@ FROM golang@sha256:f6751d823c26342f9506c03797d2527668d095b0a15f1862cddb4d927a7a4
 
 RUN apk add --no-cache git make gcc musl-dev binutils-gold
 
+# A build context sent as a tar archive keeps its files' owners, and git
+# refuses to read a checkout owned by another user. Trust this one
+# whoever owns it.
+RUN git config --system --add safe.directory /src
+
 # Force BuildKit to run the lint stage before proceeding
 COPY --from=lint /src/go.sum /dev/null
 
@@ -34,11 +41,26 @@ COPY . .
 # Run the tests - build fails if any test fails
 RUN make test
 
-# Build the binary. .dockerignore leaves out .git, so `git describe` in
+# Version stamped into the binary: the VERSION build arg when one is
-# the Makefile cannot find the version here: script/docker passes it as
+# given and not empty (script/docker passes one), otherwise what
-# --build-arg VERSION, and a build that passes none reports `dev`.
+# `git describe` says of the .git in the build context, so a plain
-ARG VERSION=dev
+# `docker build .` of a clone stamps its tag or short commit. The build
-RUN make build VERSION="${VERSION}"
+# arg reaches make through the environment.
+ARG VERSION
+
+# A context that carries .git, as a directory or as a file, must yield a
+# real version: one that is empty, `dev` or `unknown` cannot be traced
+# back to a commit.
+RUN version="$(make version)"; \
+    if [ -e .git ]; then \
+        case "$version" in \
+            "" | dev | unknown) \
+                echo "version is \"$version\" although the build context carries .git" >&2; \
+                exit 1 ;; \
+        esac; \
+    fi
+
+RUN make build
 
 # Runtime stage
 # alpine 3.21, 2026-02-28

Dockerfile.fmt

@@ -0,0 +1,54 @@
+# prettier over the markdown, in a container, so it is never installed
+# on the host. script/fmt-check-markdown builds the fmt-check stage;
+# script/fmt builds fmt-out and takes the formatted files back.
+
+# node:22-bookworm-slim, 2026-09-05
+FROM node:22-bookworm-slim@sha256:83f487e0a63425e5b4d146fb5e5be574bcbe1b7b843d3ebafdd95eaf7767a7e5 AS nodedeps
+
+# prettier lives outside /src so that a `COPY . .` of the repo cannot
+# overwrite it, and so that node_modules never appears in the tree
+# prettier is about to walk.
+WORKDIR /tools
+
+# package.json pins the version and yarn.lock pins the bytes:
+# --frozen-lockfile installs exactly the lockfile's resolution and fails
+# if package.json disagrees with it, so the tool cannot float between
+# runs. yarn is the one in the image above.
+COPY package.json yarn.lock ./
+RUN yarn install --frozen-lockfile --non-interactive --no-progress
+
+ENV PATH="/tools/node_modules/.bin:${PATH}"
+
+WORKDIR /src
+
+# Read-only markdown check. Must match $stage in
+# script/fmt-check-markdown.
+FROM nodedeps AS fmt-check
+
+COPY . .
+
+# --config, not discovery: a .prettierrc that failed to arrive would
+# otherwise leave prettier on its defaults, where proseWrap is "preserve"
+# and every wrap this check exists to enforce passes. Missing the file is
+# a hard error instead. --no-editorconfig so that .prettierrc alone sets
+# the style.
+RUN prettier --config .prettierrc --no-editorconfig --check "**/*.md"
+
+# Write path. Not a check: script/fmt builds this and takes the files.
+FROM nodedeps AS fmt
+
+COPY . .
+
+RUN prettier --config .prettierrc --no-editorconfig --write "**/*.md"
+
+# Only the markdown leaves, with its paths intact, so that the export
+# below cannot put anything else back over the caller's working tree.
+RUN mkdir -p /out && cd /src && \
+    find . -name '*.md' -type f -exec cp --parents '{}' /out/ ';'
+
+# Export target: `docker build --target fmt-out --output type=local`
+# writes /out's tree into a directory on the client, which is how
+# script/fmt gets formatted markdown back without a bind mount.
+# Must match $stage in script/fmt.
+FROM scratch AS fmt-out
+COPY --from=fmt /out/ /

Makefile

@@ -1,9 +1,13 @@
-.PHONY: all bootstrap setup build lint fmt fmt-check test check clean hooks docker
+.PHONY: all bootstrap setup build version lint fmt fmt-check test check clean hooks docker
 
 BINARY := dnswatcher
-# `make build VERSION=...` overrides this; the Dockerfile does so, as the
+# VERSION given on the command line (`make build VERSION=...`) or in the
-# image has no .git to describe.
+# environment, which is how the Dockerfile's VERSION build arg arrives,
-VERSION := $(shell git describe --tags --always --dirty 2>/dev/null || echo "dev")
+# wins over what `git describe` says of this checkout. An empty one counts
+# as not given; `override` is what replaces an empty command-line value.
+ifeq ($(VERSION),)
+override VERSION := $(shell git describe --tags --always --dirty 2>/dev/null || echo "dev")
+endif
 LDFLAGS := -X main.Version=$(VERSION)
 
 # Standard targets are thin shims; the implementations live in script/
@@ -21,6 +25,10 @@ setup:
 build:
 	go build -ldflags "$(LDFLAGS)" -o bin/$(BINARY) ./cmd/dnswatcher
 
+# Prints the version `make build` stamps; the Dockerfile checks it.
+version:
+	@echo "$(VERSION)"
+
 test:
 	@script/test
 

TESTING.md

@@ -2,44 +2,43 @@
 
 ## DNS Resolution Tests
 
-DNS is never mocked in this project, not in tests and not anywhere
+DNS is never mocked in this project, not in tests and not anywhere else; see the
-else; see the README section "No DNS mocking. Ever." Every test that
+README section "No DNS mocking. Ever." Every test that looks something up in DNS
-looks something up in DNS **MUST** query live DNS servers, never a
+**MUST** query live DNS servers, never a stand-in. Logic that works on record
-stand-in. Logic that works on record data, such as comparing or
+data, such as comparing or formatting records, may be tested on that data
-formatting records, may be tested on that data directly with no
+directly with no lookup.
-lookup.
 
 ### Rationale
 
-The resolver performs iterative resolution from root nameservers through
+The resolver performs iterative resolution from root nameservers through the
-the full delegation chain. Mocked responses cannot faithfully represent
+full delegation chain. Mocked responses cannot faithfully represent the variety
-the variety of real-world DNS behavior (truncation, referrals, glue
+of real-world DNS behavior (truncation, referrals, glue records, DNSSEC, varied
-records, DNSSEC, varied response times, EDNS, etc.). Testing against
+response times, EDNS, etc.). Testing against real servers ensures the resolver
-real servers ensures the resolver works correctly in production.
+works correctly in production.
 
 ### Constraints
 
 - Tests hit real DNS infrastructure and require network access
-- Test duration depends on network conditions; timeout tuning keeps
+- Test duration depends on network conditions; timeout tuning keeps the suite
-  the suite within the 60-second target
+  within the 60-second target
-- Query timeout is calibrated to 3× maximum antipodal RTT (~300ms)
+- Query timeout is calibrated to 3× maximum antipodal RTT (~300ms) plus
-  plus processing margin
+  processing margin
 - Root server fan-out is limited to reduce parallel query load
-- Live lookups that expect an answer go through `internal/livednstest`,
+- Live lookups that expect an answer go through `internal/livednstest`, which
-  which limits how many run at once in a test binary and retries a
+  limits how many run at once in a test binary and retries a lookup that got
-  lookup that got none
+  none
-- Flaky failures from transient network issues are acceptable and
+- Flaky failures from transient network issues are acceptable and should be
-  should be investigated as potential resolver bugs, not papered over
+  investigated as potential resolver bugs, not papered over with mocks or skip
-  with mocks or skip flags
+  flags
 
 ### What NOT to do
 
-- **Do not mock, fake or stub DNS** anywhere: no stand-in `DNSClient`,
+- **Do not mock, fake or stub DNS** anywhere: no stand-in `DNSClient`, no
-  no stand-in for the watcher's `DNSResolver`, no fake DNS server, no
+  stand-in for the watcher's `DNSResolver`, no fake DNS server, no canned
-  canned responses
+  responses
 - **Do not add `-short` flags** to skip slow tests
 - **Do not increase `-timeout`** to hide hanging queries
-- **Do not remove `-count=1` from `script/test`** — Go's test cache
+- **Do not remove `-count=1` from `script/test`** — Go's test cache replays a
-  replays a previous run's output without querying anything, so a
+  previous run's output without querying anything, so a cached pass is not
-  cached pass is not evidence that live resolution works
+  evidence that live resolution works
 - **Do not modify linter configuration** to suppress findings

TODO.md

@@ -1,12 +1,12 @@
 # Workflow
 
-* branch (from `next`)
+- branch (from `next`)
-* do the work in Next Step
+- do the work in Next Step
-* move Next Step to the top of Completed Steps
+- move Next Step to the top of Completed Steps
-* move the top item of Future Steps into Next Step
+- move the top item of Future Steps into Next Step
-* commit (`TODO.md` changes in the same commit as the work)
+- commit (`TODO.md` changes in the same commit as the work)
-* push
+- push
-* open a PR against `next`
+- open a PR against `next`
 
 # Status
 
@@ -15,11 +15,76 @@ on the 1.0 milestone: https://git.eeqj.de/sneak/dnswatcher/milestone/7
 
 # Next Step
 
-trial run of the finished image:
+trial run of the finished image: https://git.eeqj.de/sneak/dnswatcher/issues/149
-https://git.eeqj.de/sneak/dnswatcher/issues/149
 
 # Completed Steps
 
+- 2026-10-02: a nameserver whose query for one record type failed while the
+  others answered with no records is `ok`, not `nodata` (closes #253).
+- 2026-10-02: a domain that does not exist is shown so, with no nameservers; no
+  name gets a parent's nameservers when its own did not answer (closes #222).
+- 2026-10-02: the refused-query test sends one query to four operators' public
+  resolvers in turn until one replies, not eight to one operator (closes #251).
+- 2026-10-02: a name removed from `DNSWATCHER_TARGETS` leaves the state, and so
+  the dashboard and API, at startup, before the first check (closes #223).
+- 2026-10-02: a record type whose query to a nameserver fails keeps its previous
+  records and alerts nothing; the other types are still saved (closes #231).
+- 2026-10-02: a Port Change notification lists the port's domains on a
+  `Domains:` line and its hostnames on a `Hostnames:` line (closes #248).
+- 2026-10-02: the dashboard's Ports table and `/api/v1/status` port entries list
+  a port's domains apart from its hostnames (closes #245).
+- 2026-10-02: nameservers a referral names without addresses are looked up,
+  three deep at most; `pool.ntp.org`'s nameservers resolve (closes #221).
+- 2026-10-02: an apex domain is not counted or listed as a hostname; its records
+  show under Domains, and notifications about them say `Domain:` (closes #224).
+- 2026-10-02: the dashboard lists each nameserver's record types in one fixed
+  order, the README's, then any other type, not a random one (closes #226).
+- 2026-10-02: the dashboard and `/api/v1/status` show why a nameserver query or
+  a certificate check failed, which only the state file showed (closes #225).
+- 2026-10-02: a name's CNAME is stored once per nameserver, not once per record
+  type asked for; a state file with repeats loads each value once (closes #220).
+- 2026-10-02: a DNS lookup that shutdown cuts short logs no error; one that
+  fails otherwise, or runs out of time, still does (closes #229).
+- 2026-10-02: Record Change and Inconsistency notifications list only the record
+  types that differ, each with its values as plain text (closes #219).
+- 2026-10-02: the startup notification no longer says every notification
+  endpoint works; it says it is a test sent to each of them (closes #230).
+- 2026-10-02: a Mattermost webhook that answers an HTTP error is logged as
+  `mattermost notification failed`, not as a Slack failure (closes #227).
+- 2026-10-02: durations in the log are written as text such as `2m0s`, not as a
+  bare count of nanoseconds (closes #228).
+- 2026-10-02: a watched name whose nameservers answer with a CNAME and no
+  address gets port and TLS checks at the end of its CNAME chain (closes #203).
+- 2026-10-02: a resolver test that reads one record type from a nameserver's
+  answer asks again when that type is missing from it (closes #218).
+- 2026-10-02: a plain `docker build .` of a clone stamps its tag or short
+  commit, not `dev`: the build context now carries `.git` (closes #210).
+- 2026-10-02: a query a server refuses is not resent asking for recursion, and
+  every root server refusing is reported as DNS interception (closes #206).
+- 2026-10-02: a push to a branch cancels that branch's older CI run, and the
+  checkout leaves no token in `.git/config` (closes #216).
+- 2026-10-02: watcher tests send far fewer queries and a live attempt may take
+  18s; nameserver addresses are asked only for A, AAAA, CNAME (closes #214).
+- 2026-10-02: the resolver tries root servers, and every other server list it
+  walks, in a random order each time, not always from the top (closes #138).
+- 2026-10-02: a name listed more than once in `DNSWATCHER_TARGETS`, in any
+  letter case or with a trailing dot, is watched once (closes #207).
+- 2026-10-01: README checked against the code and corrected: metrics, CORS,
+  notification retries, CNAMEs, state file fields, Design tree (closes #108).
+- 2026-10-01: a certificate within the expiry warning period is warned about on
+  every TLS check, where some checks used to skip it at random (closes #204).
+- 2026-10-01: a domain's NS set is its delegation from the parent zone's
+  servers, not whichever of its own servers answered first (closes #200).
+- 2026-10-01: README has Getting Started, Rationale and TODO sections, and its
+  Architecture section is now Design, in the order policy sets (closes #173).
+- 2026-10-01: a zone's server that answers SERVFAIL or a referral leading no
+  closer is passed over for the next, as one that times out is (closes #197).
+- 2026-10-01: when none of a configured name's nameservers answered, the port
+  state saved for its addresses is kept, not removed (closes #193).
+- 2026-10-01: `ResolveIPAddresses` returns an error, not no addresses, when no
+  nameserver of the name's zone answered (closes #190).
+- 2026-10-01: `make fmt` and `make fmt-check` cover Markdown with prettier, run
+  in Docker at the version pinned by `yarn.lock` (closes #119).
 - 2026-10-01: `make fmt-check` fails on a file `goimports` would change; both
   format scripts run `goimports` at its pinned commit, not from `PATH` (#119).
 - 2026-10-01: a hostname is queried at the servers of the zone it is in, found
@@ -114,10 +179,4 @@ https://git.eeqj.de/sneak/dnswatcher/issues/149
 
 - 1.0 readiness: run it with a real config and read the logs:
   https://git.eeqj.de/sneak/dnswatcher/issues/66
-- Markdown formatting with prettier:
-  https://git.eeqj.de/sneak/dnswatcher/issues/119
-- README accuracy sweep: https://git.eeqj.de/sneak/dnswatcher/issues/108
-- README sections required by policy:
-  https://git.eeqj.de/sneak/dnswatcher/issues/173
-- fixed root server order: https://git.eeqj.de/sneak/dnswatcher/issues/138
 - review toward 1.0: https://git.eeqj.de/sneak/dnswatcher/issues/144

internal/config/classify.go

@@ -57,17 +57,22 @@ func ClassifyDNSName(name string) (DNSNameType, error) {
 
 // ClassifyTargets splits a list of DNS names into apex domains and
 // hostnames using the Public Suffix List. It returns an error if any
-// name cannot be classified.
+// name cannot be classified. A name given more than once, in any letter
+// case or with a trailing dot, is kept once.
 func ClassifyTargets(targets []string) ([]string, []string, error) {
 	var domains, hostnames []string
 
+	seen := make(map[string]bool)
+
 	for _, t := range targets {
 		normalized := strings.ToLower(strings.TrimSuffix(strings.TrimSpace(t), "."))
 
-		if normalized == "" {
+		if normalized == "" || seen[normalized] {
 			continue
 		}
 
+		seen[normalized] = true
+
 		typ, classErr := ClassifyDNSName(normalized)
 		if classErr != nil {
 			return nil, nil, classErr

internal/config/classify_test.go

@@ -1,6 +1,7 @@
 package config_test
 
 import (
+	"slices"
 	"testing"
 
 	"sneak.berlin/go/dnswatcher/internal/config"
@@ -93,6 +94,29 @@ func TestClassifyTargets(t *testing.T) {
 	}
 }
 
+func TestClassifyTargetsKeepsEachNameOnce(t *testing.T) {
+	t.Parallel()
+
+	domains, hostnames, err := config.ClassifyTargets([]string{
+		"example.org",
+		"Example.org.",
+		"www.example.org",
+		"EXAMPLE.ORG",
+		"WWW.Example.org.",
+	})
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	if !slices.Equal(domains, []string{"example.org"}) {
+		t.Errorf("domains = %v, want [example.org]", domains)
+	}
+
+	if !slices.Equal(hostnames, []string{"www.example.org"}) {
+		t.Errorf("hostnames = %v, want [www.example.org]", hostnames)
+	}
+}
+
 func TestClassifyTargetsRejectsPublicSuffix(t *testing.T) {
 	t.Parallel()
 

internal/handlers/dashboard.go

@@ -1,11 +1,14 @@
 package handlers
 
 import (
+	"cmp"
 	"embed"
 	"fmt"
 	"html/template"
+	"maps"
 	"math"
 	"net/http"
+	"slices"
 	"strings"
 	"time"
 
@@ -40,9 +43,16 @@ func newDashboardTemplate() *template.Template {
 	)
 }
 
-// dashboardData is the data passed to the dashboard template.
+// dashboardData is the data passed to the dashboard template. Hostnames
+// and DomainRecords split the records in Snapshot.Hostnames, which also
+// holds the apex domains' own (see splitHostnames). Ports holds
+// Snapshot.Ports with each port's names split into domains and
+// hostnames, as /api/v1/status gives them (see buildPorts).
 type dashboardData struct {
 	Snapshot      state.Snapshot
+	Hostnames     map[string]*state.HostnameState
+	DomainRecords map[string]*state.HostnameState
+	Ports         map[string]*statusPortInfo
 	Alerts        []notify.AlertEntry
 	StateAge      string
 	GeneratedAt   string
@@ -58,9 +68,13 @@ func (h *Handlers) HandleDashboard() http.HandlerFunc {
 	) {
 		snap := h.state.GetSnapshot()
 		alerts := h.notifyHistory.Recent()
+		hostnames, domainRecords := splitHostnames(snap)
 
 		data := dashboardData{
 			Snapshot:      snap,
+			Hostnames:     hostnames,
+			DomainRecords: domainRecords,
+			Ports:         buildPorts(snap),
 			Alerts:        alerts,
 			StateAge:      relTime(snap.LastUpdated),
 			GeneratedAt:   time.Now().UTC().Format("2006-01-02 15:04:05"),
@@ -122,16 +136,37 @@ func joinStrings(items []string, sep string) string {
 }
 
 // formatRecords formats a map of record type → values into a
-// compact display string.
+// compact display string. Record types are listed in the order the
+// README lists them, any other type after them in alphabetical order,
+// so rows of nameservers with the same records read the same.
 func formatRecords(records map[string][]string) string {
 	if len(records) == 0 {
 		return "-"
 	}
 
+	order := []string{"A", "AAAA", "CNAME", "MX", "TXT", "SRV", "CAA", "NS"}
+
+	position := func(rtype string) int {
+		i := slices.Index(order, rtype)
+		if i < 0 {
+			return len(order)
+		}
+
+		return i
+	}
+
+	rtypes := slices.Collect(maps.Keys(records))
+	slices.SortFunc(rtypes, func(a, b string) int {
+		return cmp.Or(
+			cmp.Compare(position(a), position(b)),
+			strings.Compare(a, b),
+		)
+	})
+
 	var parts []string
 
-	for rtype, values := range records {
+	for _, rtype := range rtypes {
-		for _, v := range values {
+		for _, v := range records[rtype] {
 			parts = append(parts, rtype+": "+v)
 		}
 	}

internal/handlers/dashboard_test.go

@@ -1,6 +1,8 @@
 package handlers_test
 
 import (
+	"regexp"
+	"strings"
 	"testing"
 	"time"
 
@@ -78,3 +80,194 @@ func TestFormatRecords(t *testing.T) {
 		t.Errorf("unexpected format: %q", got)
 	}
 }
+
+// TestFormatRecordsTypeOrder checks that record types are listed in
+// the README's order (A, AAAA, CNAME, MX, TXT, SRV, CAA, NS), with
+// any other type after them in alphabetical order.
+func TestFormatRecordsTypeOrder(t *testing.T) {
+	t.Parallel()
+
+	got := handlers.FormatRecords(map[string][]string{
+		"SOA":   {"ns1.example.com. hostmaster.example.com. 1 2 3 4 5"},
+		"NS":    {"ns1.example.com.", "ns2.example.com."},
+		"CAA":   {`0 issue "letsencrypt.org"`},
+		"DNAME": {"example.net."},
+		"TXT":   {"v=spf1 -all"},
+		"SRV":   {"10 5 443 www.example.com."},
+		"MX":    {"10 mail.example.com."},
+		"CNAME": {"www.example.com."},
+		"AAAA":  {"2001:db8::1"},
+		"A":     {"192.0.2.1"},
+	})
+
+	want := strings.Join([]string{
+		"A: 192.0.2.1",
+		"AAAA: 2001:db8::1",
+		"CNAME: www.example.com.",
+		"MX: 10 mail.example.com.",
+		"TXT: v=spf1 -all",
+		"SRV: 10 5 443 www.example.com.",
+		`CAA: 0 issue "letsencrypt.org"`,
+		"NS: ns1.example.com.",
+		"NS: ns2.example.com.",
+		"DNAME: example.net.",
+		"SOA: ns1.example.com. hostmaster.example.com. 1 2 3 4 5",
+	}, ", ")
+
+	if got != want {
+		t.Errorf("FormatRecords lists types out of order:\n got %q\nwant %q",
+			got, want)
+	}
+}
+
+// dashboardRow returns the table row of page that contains name.
+func dashboardRow(t *testing.T, page string, name string) string {
+	t.Helper()
+
+	for row := range strings.SplitSeq(page, "<tr") {
+		if strings.Contains(row, name) {
+			return row
+		}
+	}
+
+	t.Fatalf("dashboard has no row containing %q", name)
+
+	return ""
+}
+
+// TestDashboardShowsFailureReasons checks that the dashboard shows the
+// reason in the row of a failed nameserver and of a failed certificate,
+// and not in the row of a nameserver that answered.
+func TestDashboardShowsFailureReasons(t *testing.T) {
+	t.Parallel()
+
+	page := get(t, newHandlersWithFailures(t).HandleDashboard())
+
+	if !strings.Contains(dashboardRow(t, page, failedNS), nsFailureReason) {
+		t.Errorf("row of %s does not show %q", failedNS, nsFailureReason)
+	}
+
+	if strings.Contains(dashboardRow(t, page, answeringNS), nsFailureReason) {
+		t.Errorf("row of %s shows %q", answeringNS, nsFailureReason)
+	}
+
+	if !strings.Contains(dashboardRow(t, page, certKey), certFailedReason) {
+		t.Errorf("row of %s does not show %q", certKey, certFailedReason)
+	}
+}
+
+// dashboardSection returns the section of page under heading.
+func dashboardSection(t *testing.T, page string, heading string) string {
+	t.Helper()
+
+	for section := range strings.SplitSeq(page, "<section") {
+		words := strings.Join(strings.Fields(section), " ")
+		if strings.Contains(words, "> "+heading+" </h2>") {
+			return section
+		}
+	}
+
+	t.Fatalf("dashboard has no section headed %q", heading)
+
+	return ""
+}
+
+// TestDashboardShowsDomainRecordsUnderDomains checks that the dashboard
+// shows an apex domain's own records in the Domains section, and
+// neither lists nor counts the domain as a hostname.
+func TestDashboardShowsDomainRecordsUnderDomains(t *testing.T) {
+	t.Parallel()
+
+	page := get(t, newHandlersWithFailures(t).HandleDashboard())
+
+	domains := dashboardSection(t, page, "Domains")
+	if !strings.Contains(dashboardRow(t, domains, domainAddress), testDomain) {
+		t.Errorf("row of %s does not name %s", domainAddress, testDomain)
+	}
+
+	if strings.Contains(dashboardSection(t, page, "Hostnames"), testDomain) {
+		t.Errorf("Hostnames section lists the domain %s", testDomain)
+	}
+
+	words := strings.Join(strings.Fields(page), " ")
+
+	footer := "monitoring 2 domains + 1 hostnames"
+	if !strings.Contains(words, footer) {
+		t.Errorf("dashboard does not say %q", footer)
+	}
+
+	// With the tags taken out, the summary bar starts "Domains 2
+	// Hostnames 1".
+	text := regexp.MustCompile(`<[^>]*>`).ReplaceAllString(page, " ")
+	summary := "Domains 2 Hostnames 1"
+
+	if !strings.Contains(strings.Join(strings.Fields(text), " "), summary) {
+		t.Errorf("summary bar does not say %q", summary)
+	}
+}
+
+// TestDashboardMarksDomainThatDoesNotExist checks that the Domains
+// section says a domain that does not exist does not exist, and does
+// not say so of a domain that exists.
+func TestDashboardMarksDomainThatDoesNotExist(t *testing.T) {
+	t.Parallel()
+
+	page := get(t, newHandlersWithFailures(t).HandleDashboard())
+	domains := dashboardSection(t, page, "Domains")
+
+	if !strings.Contains(dashboardRow(t, domains, missingDomain), "does not exist") {
+		t.Errorf("row of %s does not say it does not exist", missingDomain)
+	}
+
+	if strings.Contains(dashboardRow(t, domains, testDomain), "does not exist") {
+		t.Errorf("row of %s says it does not exist", testDomain)
+	}
+}
+
+// rowCells returns the text of each cell of a dashboard table row
+// whose cells start with tag, "<th" or "<td".
+func rowCells(row string, tag string) []string {
+	tags := regexp.MustCompile(`<[^>]*>`)
+	parts := strings.Split(row, tag)[1:]
+	cells := make([]string, 0, len(parts))
+
+	for _, cell := range parts {
+		text := tags.ReplaceAllString(tag+cell, " ")
+		cells = append(cells, strings.Join(strings.Fields(text), " "))
+	}
+
+	return cells
+}
+
+// TestDashboardPortsTellDomainsFromHostnames checks that the Ports
+// table lists an apex domain under Domains and a hostname under
+// Hostnames when both resolve to the port's address.
+func TestDashboardPortsTellDomainsFromHostnames(t *testing.T) {
+	t.Parallel()
+
+	page := get(t, newHandlersWithFailures(t).HandleDashboard())
+	ports := dashboardSection(t, page, "Ports")
+
+	headings := rowCells(dashboardRow(t, ports, "Address</th>"), "<th")
+	cells := rowCells(dashboardRow(t, ports, sharedPort), "<td")
+
+	if len(cells) != len(headings) {
+		t.Fatalf("row of %s has cells %q under headings %q",
+			sharedPort, cells, headings)
+	}
+
+	under := make(map[string]string)
+	for i, heading := range headings {
+		under[heading] = cells[i]
+	}
+
+	if under["Domains"] != testDomain {
+		t.Errorf("row of %s lists %q under Domains, want %q",
+			sharedPort, under["Domains"], testDomain)
+	}
+
+	if under["Hostnames"] != testHostname {
+		t.Errorf("row of %s lists %q under Hostnames, want %q",
+			sharedPort, under["Hostnames"], testHostname)
+	}
+}

internal/handlers/status.go

@@ -9,8 +9,13 @@ import (
 )
 
 // statusDomainInfo holds status information for a monitored domain.
+// RecordsByNameserver holds the domain's own records, in the form a
+// hostname's Nameservers holds the hostname's. NXDomain is true when
+// the domain's parent zone's servers answered that it does not exist.
 type statusDomainInfo struct {
 	Nameservers         []string                         `json:"nameservers"`
+	RecordsByNameserver map[string]*statusHostnameNSInfo `json:"recordsByNameserver"`
+	NXDomain            bool                             `json:"nxdomain"`
 	LastChecked         time.Time                        `json:"lastChecked"`
 }
 
@@ -18,6 +23,7 @@ type statusDomainInfo struct {
 type statusHostnameNSInfo struct {
 	Records     map[string][]string `json:"records"`
 	Status      string              `json:"status"`
+	Error       string              `json:"error,omitempty"`
 	LastChecked time.Time           `json:"lastChecked"`
 }
 
@@ -28,8 +34,11 @@ type statusHostnameInfo struct {
 }
 
 // statusPortInfo holds status information for a monitored port.
+// Domains and Hostnames list the apex domains and the hostnames that
+// resolve to its address.
 type statusPortInfo struct {
 	Open        bool      `json:"open"`
+	Domains     []string  `json:"domains"`
 	Hostnames   []string  `json:"hostnames"`
 	LastChecked time.Time `json:"lastChecked"`
 }
@@ -41,6 +50,7 @@ type statusCertificateInfo struct {
 	NotAfter                time.Time `json:"notAfter"`
 	SubjectAlternativeNames []string  `json:"subjectAlternativeNames"`
 	Status                  string    `json:"status"`
+	Error                   string    `json:"error,omitempty"`
 	LastChecked             time.Time `json:"lastChecked"`
 }
 
@@ -94,21 +104,44 @@ func buildStatusResponse(
 		LastUpdated:  snap.LastUpdated,
 		Domains:      make(map[string]*statusDomainInfo),
 		Hostnames:    make(map[string]*statusHostnameInfo),
-		Ports:        make(map[string]*statusPortInfo),
 		Certificates: make(map[string]*statusCertificateInfo),
 	}
 
-	buildDomains(snap, resp)
+	hostnames, domainRecords := splitHostnames(snap)
-	buildHostnames(snap, resp)
+
-	buildPorts(snap, resp)
+	buildDomains(snap, domainRecords, resp)
+	buildHostnames(hostnames, resp)
+	resp.Ports = buildPorts(snap)
 	buildCertificates(snap, resp)
 	buildCounts(resp)
 
 	return resp
 }
 
+// splitHostnames returns the records saved in snap.Hostnames in two
+// maps: the hostnames' and the apex domains' own. The watcher saves a
+// domain's own records there under the domain's name, which has an
+// entry in snap.Domains too.
+func splitHostnames(
+	snap state.Snapshot,
+) (map[string]*state.HostnameState, map[string]*state.HostnameState) {
+	hostnames := make(map[string]*state.HostnameState)
+	domainRecords := make(map[string]*state.HostnameState)
+
+	for name, hs := range snap.Hostnames {
+		if _, isDomain := snap.Domains[name]; isDomain {
+			domainRecords[name] = hs
+		} else {
+			hostnames[name] = hs
+		}
+	}
+
+	return hostnames, domainRecords
+}
+
 func buildDomains(
 	snap state.Snapshot,
+	domainRecords map[string]*state.HostnameState,
 	resp *statusResponse,
 ) {
 	for name, ds := range snap.Domains {
@@ -116,22 +149,37 @@ func buildDomains(
 		copy(ns, ds.Nameservers)
 		sort.Strings(ns)
 
+		records := make(map[string]*statusHostnameNSInfo)
+		if hs, ok := domainRecords[name]; ok {
+			records = nameserverInfo(hs)
+		}
+
 		resp.Domains[name] = &statusDomainInfo{
 			Nameservers:         ns,
+			RecordsByNameserver: records,
+			NXDomain:            ds.NXDomain,
 			LastChecked:         ds.LastChecked,
 		}
 	}
 }
 
 func buildHostnames(
-	snap state.Snapshot,
+	hostnames map[string]*state.HostnameState,
 	resp *statusResponse,
 ) {
-	for name, hs := range snap.Hostnames {
+	for name, hs := range hostnames {
-		info := &statusHostnameInfo{
+		resp.Hostnames[name] = &statusHostnameInfo{
-			Nameservers: make(map[string]*statusHostnameNSInfo),
+			Nameservers: nameserverInfo(hs),
 			LastChecked: hs.LastChecked,
 		}
+	}
+}
+
+// nameserverInfo copies each nameserver's answer saved in hs.
+func nameserverInfo(
+	hs *state.HostnameState,
+) map[string]*statusHostnameNSInfo {
+	info := make(map[string]*statusHostnameNSInfo)
 
 	for ns, nsState := range hs.RecordsByNameserver {
 		recs := make(map[string][]string, len(nsState.Records))
@@ -141,32 +189,47 @@ func buildHostnames(
 			recs[rtype] = copied
 		}
 
-			info.Nameservers[ns] = &statusHostnameNSInfo{
+		info[ns] = &statusHostnameNSInfo{
 			Records:     recs,
 			Status:      nsState.Status,
+			Error:       nsState.Error,
 			LastChecked: nsState.LastChecked,
 		}
 	}
 
-		resp.Hostnames[name] = info
+	return info
-	}
 }
 
-func buildPorts(
+// buildPorts returns the port entries saved in snap. A port entry
-	snap state.Snapshot,
+// saves apex domains with its hostnames; they are told apart as in
-	resp *statusResponse,
+// splitHostnames, by a domain entry in snap.Domains.
-) {
+func buildPorts(snap state.Snapshot) map[string]*statusPortInfo {
+	ports := make(map[string]*statusPortInfo, len(snap.Ports))
+
 	for key, ps := range snap.Ports {
-		hostnames := make([]string, len(ps.Hostnames))
+		domains := []string{}
-		copy(hostnames, ps.Hostnames)
+		hostnames := []string{}
+
+		for _, name := range ps.Hostnames {
+			if _, isDomain := snap.Domains[name]; isDomain {
+				domains = append(domains, name)
+			} else {
+				hostnames = append(hostnames, name)
+			}
+		}
+
+		sort.Strings(domains)
 		sort.Strings(hostnames)
 
-		resp.Ports[key] = &statusPortInfo{
+		ports[key] = &statusPortInfo{
 			Open:        ps.Open,
+			Domains:     domains,
 			Hostnames:   hostnames,
 			LastChecked: ps.LastChecked,
 		}
 	}
+
+	return ports
 }
 
 func buildCertificates(
@@ -183,6 +246,7 @@ func buildCertificates(
 			NotAfter:                cs.NotAfter,
 			SubjectAlternativeNames: sans,
 			Status:                  cs.Status,
+			Error:                   cs.Error,
 			LastChecked:             cs.LastChecked,
 		}
 	}

internal/handlers/status_test.go

@@ -0,0 +1,309 @@
+package handlers_test
+
+import (
+	"encoding/json"
+	"net/http"
+	"net/http/httptest"
+	"slices"
+	"testing"
+	"time"
+
+	"go.uber.org/fx/fxtest"
+
+	"sneak.berlin/go/dnswatcher/internal/config"
+	"sneak.berlin/go/dnswatcher/internal/globals"
+	"sneak.berlin/go/dnswatcher/internal/handlers"
+	"sneak.berlin/go/dnswatcher/internal/logger"
+	"sneak.berlin/go/dnswatcher/internal/notify"
+	"sneak.berlin/go/dnswatcher/internal/state"
+)
+
+// The state the handler tests serve: www.example.com has one nameserver
+// that answered and one whose query failed, and its certificate check
+// failed. example.net is an apex domain, whose own records are saved
+// with the hostnames' records, as the watcher saves them. Both names
+// resolve to domainAddress, whose port 443 entry lists them.
+// missingDomain is an apex domain whose parent zone's servers answered
+// that it does not exist, saved with no nameservers and no records.
+const (
+	missingDomain    = "does-not-exist.example"
+	testHostname     = "www.example.com"
+	answeringNS      = "ns1.example.com."
+	failedNS         = "ns2.example.com."
+	nsFailureReason  = "server returned a referral"
+	certKey          = "192.0.2.1:443:www.example.com"
+	certFailedReason = "x509: certificate has expired or is not yet valid"
+	testDomain       = "example.net"
+	domainNS         = "a.iana-servers.net."
+	domainAddress    = "192.0.2.2"
+	sharedPort       = domainAddress + ":443"
+)
+
+// newHandlersWithFailures builds real Handlers whose state holds the
+// entries described above.
+func newHandlersWithFailures(t *testing.T) *handlers.Handlers {
+	t.Helper()
+
+	glob, err := globals.New(nil)
+	if err != nil {
+		t.Fatalf("globals.New: %v", err)
+	}
+
+	log, err := logger.New(nil, logger.Params{Globals: glob})
+	if err != nil {
+		t.Fatalf("logger.New: %v", err)
+	}
+
+	notifier, err := notify.New(fxtest.NewLifecycle(t), notify.Params{
+		Logger: log,
+		Config: &config.Config{},
+	})
+	if err != nil {
+		t.Fatalf("notify.New: %v", err)
+	}
+
+	st, err := state.New(fxtest.NewLifecycle(t), state.Params{
+		Logger: log,
+		Config: &config.Config{DataDir: t.TempDir()},
+	})
+	if err != nil {
+		t.Fatalf("state.New: %v", err)
+	}
+
+	setTestState(st)
+
+	hnd, err := handlers.New(nil, handlers.Params{
+		Logger:  log,
+		Globals: glob,
+		State:   st,
+		Notify:  notifier,
+	})
+	if err != nil {
+		t.Fatalf("handlers.New: %v", err)
+	}
+
+	return hnd
+}
+
+// setTestState sets the entries described above in st.
+func setTestState(st *state.State) {
+	now := time.Now()
+
+	st.SetHostnameState(testHostname, &state.HostnameState{
+		RecordsByNameserver: map[string]*state.NameserverRecordState{
+			answeringNS: {
+				Records: map[string][]string{
+					"A": {"192.0.2.1", domainAddress},
+				},
+				Status:      "ok",
+				LastChecked: now,
+			},
+			failedNS: {
+				Records:     map[string][]string{},
+				Status:      "error",
+				Error:       nsFailureReason,
+				LastChecked: now,
+			},
+		},
+		LastChecked: now,
+	})
+
+	st.SetCertificateState(certKey, &state.CertificateState{
+		Status:      "error",
+		Error:       certFailedReason,
+		LastChecked: now,
+	})
+
+	st.SetDomainState(testDomain, &state.DomainState{
+		Nameservers: []string{domainNS},
+		LastChecked: now,
+	})
+
+	st.SetHostnameState(testDomain, &state.HostnameState{
+		RecordsByNameserver: map[string]*state.NameserverRecordState{
+			domainNS: {
+				Records:     map[string][]string{"A": {domainAddress}},
+				Status:      "ok",
+				LastChecked: now,
+			},
+		},
+		LastChecked: now,
+	})
+
+	st.SetPortState(sharedPort, &state.PortState{
+		Open:        true,
+		Hostnames:   []string{testDomain, testHostname},
+		LastChecked: now,
+	})
+
+	st.SetDomainState(missingDomain, &state.DomainState{
+		Nameservers: []string{},
+		NXDomain:    true,
+		LastChecked: now,
+	})
+}
+
+// get serves one GET request to handler and returns the response body.
+func get(t *testing.T, handler http.HandlerFunc) string {
+	t.Helper()
+
+	rec := httptest.NewRecorder()
+	req := httptest.NewRequestWithContext(
+		t.Context(), http.MethodGet, "/", nil,
+	)
+
+	handler(rec, req)
+
+	if rec.Code != http.StatusOK {
+		t.Fatalf("status = %d, want 200", rec.Code)
+	}
+
+	return rec.Body.String()
+}
+
+// TestStatusGivesFailureReasons checks that /api/v1/status gives the
+// reason for a failed nameserver entry and a failed certificate entry,
+// and no error for a nameserver that answered.
+func TestStatusGivesFailureReasons(t *testing.T) {
+	t.Parallel()
+
+	body := get(t, newHandlersWithFailures(t).HandleStatus())
+
+	var resp struct {
+		Hostnames map[string]struct {
+			Nameservers map[string]map[string]any `json:"nameservers"`
+		} `json:"hostnames"`
+		Certificates map[string]map[string]any `json:"certificates"`
+	}
+
+	err := json.Unmarshal([]byte(body), &resp)
+	if err != nil {
+		t.Fatalf("decoding response: %v", err)
+	}
+
+	nameservers := resp.Hostnames[testHostname].Nameservers
+
+	got := nameservers[failedNS]["error"]
+	if got != nsFailureReason {
+		t.Errorf("failed nameserver error = %v, want %q",
+			got, nsFailureReason)
+	}
+
+	_, has := nameservers[answeringNS]["error"]
+	if has {
+		t.Errorf("answering nameserver has an error field: %v",
+			nameservers[answeringNS])
+	}
+
+	got = resp.Certificates[certKey]["error"]
+	if got != certFailedReason {
+		t.Errorf("failed certificate error = %v, want %q",
+			got, certFailedReason)
+	}
+}
+
+// TestStatusGivesDomainRecordsUnderTheDomain checks that /api/v1/status
+// gives an apex domain's own records in its domain entry, and neither
+// lists nor counts the domain as a hostname.
+func TestStatusGivesDomainRecordsUnderTheDomain(t *testing.T) {
+	t.Parallel()
+
+	body := get(t, newHandlersWithFailures(t).HandleStatus())
+
+	var resp struct {
+		Counts struct {
+			Hostnames int `json:"hostnames"`
+		} `json:"counts"`
+		Domains map[string]struct {
+			RecordsByNameserver map[string]struct {
+				Records map[string][]string `json:"records"`
+			} `json:"recordsByNameserver"`
+		} `json:"domains"`
+		Hostnames map[string]any `json:"hostnames"`
+	}
+
+	err := json.Unmarshal([]byte(body), &resp)
+	if err != nil {
+		t.Fatalf("decoding response: %v", err)
+	}
+
+	if resp.Counts.Hostnames != 1 {
+		t.Errorf("counts.hostnames = %d, want 1", resp.Counts.Hostnames)
+	}
+
+	if _, listed := resp.Hostnames[testDomain]; listed {
+		t.Errorf("hostnames lists the domain %s", testDomain)
+	}
+
+	records := resp.Domains[testDomain].RecordsByNameserver[domainNS].Records
+	if !slices.Equal(records["A"], []string{domainAddress}) {
+		t.Errorf("domain %s records at %s = %v, want A %s",
+			testDomain, domainNS, records, domainAddress)
+	}
+}
+
+// TestStatusMarksDomainThatDoesNotExist checks that /api/v1/status sets
+// nxdomain for a domain that does not exist, with no nameservers or
+// records, and not for a domain that exists.
+func TestStatusMarksDomainThatDoesNotExist(t *testing.T) {
+	t.Parallel()
+
+	body := get(t, newHandlersWithFailures(t).HandleStatus())
+
+	var resp struct {
+		Domains map[string]struct {
+			Nameservers         []string       `json:"nameservers"`
+			RecordsByNameserver map[string]any `json:"recordsByNameserver"`
+			NXDomain            bool           `json:"nxdomain"`
+		} `json:"domains"`
+	}
+
+	err := json.Unmarshal([]byte(body), &resp)
+	if err != nil {
+		t.Fatalf("decoding response: %v", err)
+	}
+
+	missing := resp.Domains[missingDomain]
+	if !missing.NXDomain || len(missing.Nameservers) != 0 ||
+		len(missing.RecordsByNameserver) != 0 {
+		t.Errorf("domain %s = %+v, want nxdomain and nothing else",
+			missingDomain, missing)
+	}
+
+	if resp.Domains[testDomain].NXDomain {
+		t.Errorf("domain %s has nxdomain set", testDomain)
+	}
+}
+
+// TestStatusPortsTellDomainsFromHostnames checks that a port entry in
+// /api/v1/status lists an apex domain in domains and a hostname in
+// hostnames when both resolve to its address.
+func TestStatusPortsTellDomainsFromHostnames(t *testing.T) {
+	t.Parallel()
+
+	body := get(t, newHandlersWithFailures(t).HandleStatus())
+
+	var resp struct {
+		Ports map[string]struct {
+			Domains   []string `json:"domains"`
+			Hostnames []string `json:"hostnames"`
+		} `json:"ports"`
+	}
+
+	err := json.Unmarshal([]byte(body), &resp)
+	if err != nil {
+		t.Fatalf("decoding response: %v", err)
+	}
+
+	port := resp.Ports[sharedPort]
+
+	if !slices.Equal(port.Domains, []string{testDomain}) {
+		t.Errorf("port %s domains = %v, want [%s]",
+			sharedPort, port.Domains, testDomain)
+	}
+
+	if !slices.Equal(port.Hostnames, []string{testHostname}) {
+		t.Errorf("port %s hostnames = %v, want [%s]",
+			sharedPort, port.Hostnames, testHostname)
+	}
+}

internal/handlers/templates/dashboard.html

@@ -39,7 +39,7 @@
             Hostnames
           </div>
           <div class="text-2xl font-bold text-teal-400 mt-1">
-            {{ len .Snapshot.Hostnames }}
+            {{ len .Hostnames }}
           </div>
         </div>
         <div class="bg-surface-800 border border-slate-700/50 rounded-lg p-4">
@@ -84,7 +84,11 @@
                   {{ $name }}
                 </td>
                 <td class="py-2 px-3 text-slate-400 break-all">
+                  {{ if $ds.NXDomain }}
+                  <span class="text-red-400">does not exist</span>
+                  {{ else }}
                   {{ joinStrings $ds.Nameservers ", " }}
+                  {{ end }}
                 </td>
                 <td class="py-2 px-3 text-slate-500 whitespace-nowrap">
                   {{ relTime $ds.LastChecked }}
@@ -94,6 +98,24 @@
             </tbody>
           </table>
         </div>
+        {{ if .DomainRecords }}
+        <div class="overflow-x-auto mt-4">
+          <table class="w-full text-left text-xs">
+            <thead>
+              <tr class="text-slate-500 uppercase tracking-wider">
+                <th class="py-2 px-3">Domain</th>
+                <th class="py-2 px-3">NS</th>
+                <th class="py-2 px-3">Status</th>
+                <th class="py-2 px-3">Records</th>
+                <th class="py-2 px-3">Checked</th>
+              </tr>
+            </thead>
+            <tbody class="divide-y divide-slate-800">
+              {{ template "records" .DomainRecords }}
+            </tbody>
+          </table>
+        </div>
+        {{ end }}
         {{ else }}
         <p class="text-slate-600 italic text-xs">
           No domains configured.
@@ -108,7 +130,7 @@
         >
           Hostnames
         </h2>
-        {{ if .Snapshot.Hostnames }}
+        {{ if .Hostnames }}
         <div class="overflow-x-auto">
           <table class="w-full text-left text-xs">
             <thead>
@@ -121,39 +143,7 @@
               </tr>
             </thead>
             <tbody class="divide-y divide-slate-800">
-              {{ range $name, $hs := .Snapshot.Hostnames }}
+              {{ template "records" .Hostnames }}
-              {{ range $ns, $nsr := $hs.RecordsByNameserver }}
-              <tr class="hover:bg-surface-800/50">
-                <td class="py-2 px-3 text-slate-200 font-medium">
-                  {{ $name }}
-                </td>
-                <td class="py-2 px-3 text-slate-400 break-all">
-                  {{ $ns }}
-                </td>
-                <td class="py-2 px-3">
-                  {{ if eq $nsr.Status "ok" }}
-                  <span
-                    class="inline-block px-1.5 py-0.5 rounded text-[10px] font-bold uppercase bg-teal-900/50 text-teal-400 border border-teal-700/30"
-                    >ok</span
-                  >
-                  {{ else }}
-                  <span
-                    class="inline-block px-1.5 py-0.5 rounded text-[10px] font-bold uppercase bg-red-900/50 text-red-400 border border-red-700/30"
-                    >{{ $nsr.Status }}</span
-                  >
-                  {{ end }}
-                </td>
-                <td
-                  class="py-2 px-3 text-slate-400 break-all max-w-xs"
-                >
-                  {{ formatRecords $nsr.Records }}
-                </td>
-                <td class="py-2 px-3 text-slate-500 whitespace-nowrap">
-                  {{ relTime $nsr.LastChecked }}
-                </td>
-              </tr>
-              {{ end }}
-              {{ end }}
             </tbody>
           </table>
         </div>
@@ -171,19 +161,20 @@
         >
           Ports
         </h2>
-        {{ if .Snapshot.Ports }}
+        {{ if .Ports }}
         <div class="overflow-x-auto">
           <table class="w-full text-left text-xs">
             <thead>
               <tr class="text-slate-500 uppercase tracking-wider">
                 <th class="py-2 px-3">Address</th>
                 <th class="py-2 px-3">State</th>
+                <th class="py-2 px-3">Domains</th>
                 <th class="py-2 px-3">Hostnames</th>
                 <th class="py-2 px-3">Checked</th>
               </tr>
             </thead>
             <tbody class="divide-y divide-slate-800">
-              {{ range $key, $ps := .Snapshot.Ports }}
+              {{ range $key, $ps := .Ports }}
               <tr class="hover:bg-surface-800/50">
                 <td class="py-2 px-3 text-slate-200 font-medium">
                   {{ $key }}
@@ -201,6 +192,9 @@
                   >
                   {{ end }}
                 </td>
+                <td class="py-2 px-3 text-slate-400 break-all">
+                  {{ joinStrings $ps.Domains ", " }}
+                </td>
                 <td class="py-2 px-3 text-slate-400 break-all">
                   {{ joinStrings $ps.Hostnames ", " }}
                 </td>
@@ -258,6 +252,11 @@
                   >
                   {{ end }}
                 </td>
+                {{ if $cs.Error }}
+                <td colspan="3" class="py-2 px-3 text-red-400 break-all">
+                  <div class="max-w-xs">{{ $cs.Error }}</div>
+                </td>
+                {{ else }}
                 <td class="py-2 px-3 text-slate-200">
                   {{ $cs.CommonName }}
                 </td>
@@ -285,6 +284,7 @@
                     {{ end }}
                   {{ end }}
                 </td>
+                {{ end }}
                 <td class="py-2 px-3 text-slate-500 whitespace-nowrap">
                   {{ relTime $cs.LastChecked }}
                 </td>
@@ -363,8 +363,48 @@
         class="text-[11px] text-slate-700 border-t border-slate-800 pt-4 mt-8"
       >
         dnswatcher &middot; monitoring {{ len .Snapshot.Domains }} domains +
-        {{ len .Snapshot.Hostnames }} hostnames
+        {{ len .Hostnames }} hostnames
       </div>
     </div>
   </body>
 </html>
+{{/* ---- One row per nameserver of each name in the map it is given ---- */}}
+{{ define "records" }}
+{{ range $name, $hs := . }}
+{{ range $ns, $nsr := $hs.RecordsByNameserver }}
+<tr class="hover:bg-surface-800/50">
+  <td class="py-2 px-3 text-slate-200 font-medium">
+    {{ $name }}
+  </td>
+  <td class="py-2 px-3 text-slate-400 break-all">
+    {{ $ns }}
+  </td>
+  <td class="py-2 px-3">
+    {{ if eq $nsr.Status "ok" }}
+    <span
+      class="inline-block px-1.5 py-0.5 rounded text-[10px] font-bold uppercase bg-teal-900/50 text-teal-400 border border-teal-700/30"
+      >ok</span
+    >
+    {{ else }}
+    <span
+      class="inline-block px-1.5 py-0.5 rounded text-[10px] font-bold uppercase bg-red-900/50 text-red-400 border border-red-700/30"
+      >{{ $nsr.Status }}</span
+    >
+    {{ end }}
+  </td>
+  <td
+    class="py-2 px-3 text-slate-400 break-all max-w-xs"
+  >
+    {{ if $nsr.Error }}
+    <span class="text-red-400">{{ $nsr.Error }}</span>
+    {{ else }}
+    {{ formatRecords $nsr.Records }}
+    {{ end }}
+  </td>
+  <td class="py-2 px-3 text-slate-500 whitespace-nowrap">
+    {{ relTime $nsr.LastChecked }}
+  </td>
+</tr>
+{{ end }}
+{{ end }}
+{{ end }}

internal/livednstest/livednstest.go

@@ -9,11 +9,11 @@
 //
 //  1. Bounded concurrency. Tests run in parallel and the build hosts
 //     have many cores, so without a limit every test starts its own
-//     iterative resolution at the same instant and they all hit the
+//     iterative resolution at the same instant and they all send their
-//     first root server within a few milliseconds of each other. Root
+//     first queries to the root servers within a few milliseconds of
-//     servers rate-limit that, which shows up as a different arbitrary
+//     each other. Root servers rate-limit that, which shows up as a
-//     subset of tests failing on each run. Run caps how many live
+//     different arbitrary subset of tests failing on each run. Run caps
-//     operations are in flight at once in one test binary.
+//     how many live operations are in flight at once in one test binary.
 //
 //  2. Retry with exponential backoff. Each live operation gets several
 //     attempts with its own timeout. An attempt is retried when it
@@ -36,11 +36,18 @@ const (
 	// before the test fails.
 	attempts = 3
 
-	// AttemptTimeout bounds one attempt. Worst case for an operation
+	// AttemptTimeout bounds one attempt. It must fit the longest
-	// is attempts * AttemptTimeout plus the backoff — about 26
+	// operation, a watcher check, which sends over a hundred queries one
-	// seconds, well inside the 90-second `go test -timeout` backstop
+	// after another and on a slow build host takes several times as long
-	// even when several operations exhaust their attempts.
+	// as the few seconds it takes on a fast one. An operation whose
-	AttemptTimeout = 8 * time.Second
+	// every attempt fails takes attempts * AttemptTimeout plus the
+	// backoff, about 56 seconds, after it waits for one of the
+	// Concurrency slots that every live operation in the test binary
+	// shares. So when live DNS does not answer at all, a test binary
+	// with more live operations than slots runs into the 90-second
+	// `go test -timeout` backstop instead of each test failing on its
+	// own.
+	AttemptTimeout = 18 * time.Second
 
 	// backoffBase is the delay after the first failed attempt; it is
 	// multiplied by backoffFactor each time.

internal/notify/delivery_test.go

@@ -6,9 +6,11 @@ import (
 	"encoding/json"
 	"errors"
 	"io"
+	"maps"
 	"net/http"
 	"net/http/httptest"
 	"net/url"
+	"strings"
 	"sync"
 	"testing"
 	"time"
@@ -413,7 +415,8 @@ func sendSlackInfo(
 	svc *notify.Service, target *url.URL,
 ) error {
 	return svc.SendSlack(
-		context.Background(), target, "t", "m", prioInfo,
+		context.Background(), target, notify.ErrSlackFailed,
+		"t", "m", prioInfo,
 	)
 }
 
@@ -506,6 +509,7 @@ func TestSendSlackPayloadFields(t *testing.T) {
 	err := svc.SendSlack(
 		context.Background(),
 		webhookURL,
+		notify.ErrSlackFailed,
 		"Alert Title",
 		"Alert body text",
 		"warning",
@@ -608,7 +612,8 @@ func TestSendSlackAllColors(t *testing.T) {
 
 			err := svc.SendSlack(
 				context.Background(),
-				webhookURL, "t", "m", tc.priority,
+				webhookURL, notify.ErrSlackFailed,
+				"t", "m", tc.priority,
 			)
 			if err != nil {
 				t.Fatalf("SendSlack error: %v", err)
@@ -659,7 +664,8 @@ func TestSendSlackNetworkError(t *testing.T) {
 	)
 
 	err := svc.SendSlack(
-		context.Background(), webhookURL, "t", "m", "info",
+		context.Background(), webhookURL, notify.ErrSlackFailed,
+		"t", "m", "info",
 	)
 	if err == nil {
 		t.Fatal("expected error for network failure")
@@ -1028,6 +1034,66 @@ func TestSendNotificationMattermostError(t *testing.T) {
 	)
 }
 
+// TestSendNotificationErrorNamesEndpoint verifies that, with both
+// Slack and Mattermost set, a failed delivery's logged error names
+// the endpoint that failed. Both are sent by the Slack sender.
+func TestSendNotificationErrorNamesEndpoint(t *testing.T) {
+	t.Parallel()
+
+	srv := httptest.NewServer(
+		http.HandlerFunc(
+			func(w http.ResponseWriter, _ *http.Request) {
+				w.WriteHeader(http.StatusServiceUnavailable)
+			}),
+	)
+	defer srv.Close()
+
+	target, _ := url.Parse(srv.URL)
+
+	svc, logs := newLoggingService(http.DefaultTransport)
+	svc.SetSlackWebhookURL(target)
+	svc.SetMattermostWebhookURL(target)
+	svc.SetSleepFunc(instantSleep)
+	svc.SetRetryConfig(notify.RetryConfig{
+		MaxRetries: 1,
+		BaseDelay:  time.Millisecond,
+		MaxDelay:   time.Millisecond,
+	})
+
+	svc.SendNotification(
+		context.Background(), "t", "m", prioError,
+	)
+
+	waitForCondition(t, func() bool {
+		return svc.OutstandingDeliveries() == 0
+	})
+
+	got := map[string]string{}
+
+	for line := range strings.Lines(logs.String()) {
+		var record struct {
+			Msg      string `json:"msg"`
+			Endpoint string `json:"endpoint"`
+			Error    string `json:"error"`
+		}
+
+		_ = json.Unmarshal([]byte(line), &record)
+
+		if record.Msg == "failed to send notification after retries" {
+			got[record.Endpoint] = record.Error
+		}
+	}
+
+	want := map[string]string{
+		"slack":      "slack notification failed: status 503",
+		"mattermost": "mattermost notification failed: status 503",
+	}
+
+	if !maps.Equal(got, want) {
+		t.Errorf("logged errors = %v, want %v", got, want)
+	}
+}
+
 // ── SlackPayload JSON marshaling ──────────────────────────
 
 func TestSlackPayloadJSON(t *testing.T) {

internal/notify/export_test.go

@@ -85,10 +85,11 @@ func (svc *Service) SendNtfy(
 func (svc *Service) SendSlack(
 	ctx context.Context,
 	webhookURL *url.URL,
+	failed error,
 	title, message, priority string,
 ) error {
 	return svc.sendSlack(
-		ctx, webhookURL, title, message, priority,
+		ctx, webhookURL, failed, title, message, priority,
 	)
 }
 

internal/notify/notify.go

@@ -277,7 +277,8 @@ func (svc *Service) dispatchSlack(
 
 	svc.dispatch(ctx, "slack", func(c context.Context) error {
 		return svc.sendSlack(
-			c, svc.slackWebhookURL, title, message, priority,
+			c, svc.slackWebhookURL, ErrSlackFailed,
+			title, message, priority,
 		)
 	})
 }
@@ -294,7 +295,7 @@ func (svc *Service) dispatchMattermost(
 		ctx, "mattermost",
 		func(c context.Context) error {
 			return svc.sendSlack(
-				c, svc.mattermostWebhookURL,
+				c, svc.mattermostWebhookURL, ErrMattermostFailed,
 				title, message, priority,
 			)
 		},
@@ -370,9 +371,14 @@ type SlackAttachment struct {
 	Text  string `json:"text"`
 }
 
+// sendSlack posts to a Slack or Mattermost incoming webhook, which
+// take the same payload. An HTTP error status is returned wrapped in
+// failed, ErrSlackFailed or ErrMattermostFailed, so the error names
+// the endpoint.
 func (svc *Service) sendSlack(
 	ctx context.Context,
 	webhookURL *url.URL,
+	failed error,
 	title, message, priority string,
 ) error {
 	ctx, cancel := context.WithTimeout(
@@ -420,7 +426,7 @@ func (svc *Service) sendSlack(
 	if resp.StatusCode >= httpStatusClientError {
 		return fmt.Errorf(
 			"%w: status %d",
-			ErrSlackFailed, resp.StatusCode,
+			failed, resp.StatusCode,
 		)
 	}
 

internal/notify/retry.go

@@ -115,7 +115,9 @@ func (svc *Service) deliverWithRetry(
 			"endpoint", endpoint,
 			"attempt", attempt+1,
 			"maxAttempts", cfg.MaxRetries+1,
-			"retryIn", delay,
+			// As text: the JSON log writes a time.Duration as
+			// bare nanoseconds.
+			"retryIn", delay.String(),
 			"error", lastErr,
 		)
 

internal/notify/retry_test.go

@@ -2,6 +2,7 @@ package notify_test
 
 import (
 	"context"
+	"encoding/json"
 	"errors"
 	"net/http"
 	"net/http/httptest"
@@ -189,6 +190,50 @@ func TestDeliverWithRetryExhaustsAttempts(t *testing.T) {
 	}
 }
 
+// TestDeliverWithRetryLogsRetryInAsText checks that the wait
+// before a retry is logged as text such as "1.02s", not as a
+// count of nanoseconds.
+func TestDeliverWithRetryLogsRetryInAsText(t *testing.T) {
+	t.Parallel()
+
+	svc, logs := newLoggingService(http.DefaultTransport)
+	svc.SetRetryConfig(notify.RetryConfig{
+		MaxRetries: 1,
+		BaseDelay:  time.Second,
+		MaxDelay:   time.Second,
+	})
+
+	var waited time.Duration
+
+	svc.SetSleepFunc(func(d time.Duration) <-chan time.Time {
+		waited = d
+
+		return instantSleep(d)
+	})
+
+	_ = svc.DeliverWithRetry(
+		context.Background(), "test",
+		func(_ context.Context) error {
+			return errFail
+		},
+	)
+
+	// With one retry, only the first failure is logged.
+	var record map[string]any
+
+	err := json.Unmarshal([]byte(logs.String()), &record)
+	if err != nil {
+		t.Fatalf("log is not one JSON record: %v\n%s", err, logs)
+	}
+
+	if record["retryIn"] != waited.String() {
+		t.Errorf(
+			"retryIn logged as %v, want %q",
+			record["retryIn"], waited.String(),
+		)
+	}
+}
+
 func TestDeliverWithRetryRespectsContextCancellation(
 	t *testing.T,
 ) {

internal/portcheck/portcheck.go

@@ -193,7 +193,9 @@ func (c *Checker) checkConnection(
 	c.log.Debug(
 		"port check succeeded",
 		"target", target,
-		"latency", latency,
+		// As text: the JSON log writes a time.Duration as bare
+		// nanoseconds.
+		"latency", latency.String(),
 	)
 
 	return &PortResult{

internal/resolver/errors.go

@@ -10,12 +10,46 @@ var (
 		"no authoritative nameservers found",
 	)
 
+	// ErrNXDomain is returned when the servers of the zone a domain
+	// is in answer NXDOMAIN: the domain does not exist.
+	ErrNXDomain = errors.New("domain does not exist")
+
+	// ErrNoNameserverAnswered is returned when every nameserver
+	// asked about a name timed out, failed or returned a referral,
+	// so whether the name has addresses is unknown.
+	ErrNoNameserverAnswered = errors.New("no nameserver answered")
+
+	// ErrUnusableReply is returned when a server replied with an
+	// error such as SERVFAIL, or with a referral that leads no
+	// closer to the name asked about.
+	ErrUnusableReply = errors.New(
+		"reply is an error or a referral that leads no closer",
+	)
+
+	// ErrTruncated is the reason given for a reply too large for UDP
+	// whose retry over TCP failed.
+	ErrTruncated = errors.New(
+		"reply truncated and its retry over TCP failed",
+	)
+
+	// ErrIntercepted is returned when every root server refused a
+	// query. Root servers refuse no query, so the refusals came from
+	// something on the network answering in their place.
+	ErrIntercepted = errors.New("this network intercepts DNS queries")
+
 	// ErrCNAMEDepthExceeded is returned when a CNAME chain
 	// exceeds MaxCNAMEDepth.
 	ErrCNAMEDepthExceeded = errors.New(
 		"CNAME chain depth exceeded",
 	)
 
+	// ErrLookupDepthExceeded is returned when nameserver addresses
+	// were not looked up because lookups were already maxLookupDepth
+	// deep, one inside another.
+	ErrLookupDepthExceeded = errors.New(
+		"lookups of nameserver addresses go too deep",
+	)
+
 	// ErrContextCanceled wraps context cancellation for the
 	// resolver's iterative queries.
 	ErrContextCanceled = errors.New("context canceled")

internal/resolver/export_test.go

@@ -2,20 +2,145 @@ package resolver
 
 import (
 	"context"
+	"log/slog"
+	"time"
 
 	"github.com/miekg/dns"
 )
 
+// NewWithFailingTCP returns a Resolver whose TCP client gives up before
+// it can connect, so the retry over TCP of every truncated reply fails.
+func NewWithFailingTCP(log *slog.Logger) *Resolver {
+	r := NewFromLogger(log)
+	r.tcp = &tcpClient{timeout: time.Nanosecond}
+
+	return r
+}
+
+// NewWithQueryTimeout returns a Resolver whose queries over UDP give up
+// after timeout, so a test that asks an address where nothing answers
+// does not wait out the usual timeout.
+func NewWithQueryTimeout(log *slog.Logger, timeout time.Duration) *Resolver {
+	r := NewFromLogger(log)
+	r.client = &udpClient{timeout: timeout}
+
+	return r
+}
+
+// FollowDelegation exports followDelegation for testing.
+func (r *Resolver) FollowDelegation(
+	ctx context.Context,
+	domain string,
+	servers []string,
+) ([]string, error) {
+	return r.followDelegation(ctx, domain, servers)
+}
+
+// FindAuthoritativeNameserversFrom exports findAuthoritativeNameservers
+// for testing.
+func (r *Resolver) FindAuthoritativeNameserversFrom(
+	ctx context.Context,
+	domain string,
+	servers []string,
+) ([]string, error) {
+	return r.findAuthoritativeNameservers(ctx, domain, servers)
+}
+
+// ResolveNSIterative exports resolveNSIterative for testing.
+func (r *Resolver) ResolveNSIterative(
+	ctx context.Context,
+	domain string,
+) ([]string, error) {
+	return r.resolveNSIterative(ctx, domain)
+}
+
 // ExtractRecordValue exports extractRecordValue for testing.
 func ExtractRecordValue(rr dns.RR) string {
 	return extractRecordValue(rr)
 }
 
+// CollectAnswerRecords exports collectAnswerRecords for testing.
+func CollectAnswerRecords(msg *dns.Msg, resp *NameserverResponse) {
+	var state queryState
+
+	collectAnswerRecords(msg, resp, &state)
+}
+
+// UsableReply exports usableReply for testing.
+func UsableReply(resp *dns.Msg, zone string, name string) bool {
+	return usableReply(resp, zone, name)
+}
+
+// NSSetFrom exports nsSetFrom for testing.
+func NSSetFrom(resp *dns.Msg, domain string) []string {
+	return nsSetFrom(resp, domain)
+}
+
+// CollectIPs exports collectIPs for testing.
+func CollectIPs(
+	results map[string]*NameserverResponse,
+) ([]string, string, error) {
+	return collectIPs(results)
+}
+
+// QueryServers exports queryServers for testing.
+func (r *Resolver) QueryServers(
+	ctx context.Context,
+	servers []string,
+	zone string,
+	name string,
+	qtype uint16,
+) (*dns.Msg, error) {
+	return r.queryServers(ctx, servers, zone, name, qtype)
+}
+
 // QueryEachNS exports queryEachNS for testing.
 func (r *Resolver) QueryEachNS(
 	ctx context.Context,
 	nameservers []string,
 	hostname string,
 ) (map[string]*NameserverResponse, error) {
-	return r.queryEachNS(ctx, nameservers, hostname)
+	return r.queryEachNS(ctx, nameservers, hostname, recordTypes())
+}
+
+// ResolveNSIPs exports resolveNSIPs for testing, looking each name up
+// as a lookup that no other lookup started.
+func (r *Resolver) ResolveNSIPs(
+	ctx context.Context,
+	nsNames []string,
+) []string {
+	ips, _ := r.resolveNSIPs(ctx, nsNames, 1)
+
+	return ips
+}
+
+// MaxLookupDepth exports maxLookupDepth for testing.
+const MaxLookupDepth = maxLookupDepth
+
+// QueryZone exports queryZone for testing.
+func (r *Resolver) QueryZone(
+	ctx context.Context,
+	given []string,
+	withoutAddresses []string,
+	zone string,
+	name string,
+	qtype uint16,
+	depth int,
+) (*dns.Msg, error) {
+	return r.queryZone(
+		ctx, given, withoutAddresses, zone, name, qtype, depth,
+	)
+}
+
+// RootServerList exports rootServerList for testing.
+func RootServerList() []string {
+	return rootServerList()
+}
+
+// Shuffled exports shuffled for testing.
+func Shuffled(
+	servers []string,
+	shuffle func(n int, swap func(i, j int)),
+) []string {
+	return shuffled(servers, shuffle)
 }

internal/resolver/iterative.go

@@ -4,8 +4,11 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"math/rand/v2"
 	"net"
+	"slices"
 	"sort"
+	"strconv"
 	"strings"
 	"time"
 
@@ -17,6 +20,16 @@ const (
 	maxRetries           = 2
 	maxDelegation        = 20
 	timeoutMultiplier    = 2
+
+	// maxLookupDepth is how many lookups of nameserver addresses may be
+	// under way one inside another. Looking up a nameserver's address
+	// can meet a referral that names nameservers without their
+	// addresses, which are then looked up in turn; without a limit,
+	// delegations that point at each other would never end. Each level
+	// multiplies the queries sent. pool.ntp.org needs three: the
+	// address of its nameserver g.ntpns.org can need a.ntpns.org's,
+	// which needs a bitnames.com nameserver's.
+	maxLookupDepth = 3
 )
 
 // ErrRefused is returned when a DNS server refuses a query.
@@ -87,6 +100,9 @@ func (r *Resolver) tryExchange(
 	return resp, err
 }
 
+// retryTCP returns the reply to msg over TCP when resp, its reply over
+// UDP, is truncated. When that fails it returns resp, still truncated,
+// which holds only the records that fit.
 func (r *Resolver) retryTCP(
 	ctx context.Context,
 	msg *dns.Msg,
@@ -105,9 +121,8 @@ func (r *Resolver) retryTCP(
 	return resp
 }
 
-// queryDNS sends a DNS query to a specific server IP.
+// queryDNS sends a DNS query to a specific server IP, never asking it
-// Tries non-recursive first, falls back to recursive on
+// for recursion. A reply of REFUSED is returned as ErrRefused.
-// REFUSED (handles DNS interception environments).
 func (r *Resolver) queryDNS(
 	ctx context.Context,
 	serverIP string,
@@ -130,26 +145,13 @@ func (r *Resolver) queryDNS(
 		return nil, fmt.Errorf("query %s @%s: %w", name, serverIP, err)
 	}
 
-	if resp.Rcode == dns.RcodeRefused {
-		msg.RecursionDesired = true
-
-		resp, err = r.tryExchange(ctx, msg, addr)
-		if err != nil {
-			return nil, fmt.Errorf(
-				"query %s @%s: %w", name, serverIP, err,
-			)
-		}
-
 	if resp.Rcode == dns.RcodeRefused {
 		return nil, fmt.Errorf(
 			"query %s @%s: %w", name, serverIP, ErrRefused,
 		)
 	}
-	}
 
-	resp = r.retryTCP(ctx, msg, addr, resp)
+	return r.retryTCP(ctx, msg, addr, resp), nil
-
-	return resp, nil
 }
 
 func extractNSSet(rrs []dns.RR) []string {
@@ -202,35 +204,51 @@ func glueIPs(nsNames []string, glue map[string][]net.IP) []string {
 	return ips
 }
 
+// followDelegation follows referrals from servers, the root servers, to
+// domain and returns the NS set of domain's delegation. When the servers
+// of the zone domain is in answer that domain does not exist, the error
+// is ErrNXDomain. When they answer that it has no delegation of its own,
+// because it is not the zone's apex, the set is empty and there is no
+// error. Any other error means that no such answer came.
 func (r *Resolver) followDelegation(
 	ctx context.Context,
 	domain string,
 	servers []string,
 ) ([]string, error) {
+	// servers are the root servers, the servers of zone ".".
+	zone := "."
+
+	var withoutAddresses []string
+
 	for range maxDelegation {
 		if checkCtx(ctx) != nil {
 			return nil, ErrContextCanceled
 		}
 
-		resp, err := r.queryServers(
+		resp, err := r.queryZone(
-			ctx, servers, domain, dns.TypeNS,
+			ctx, servers, withoutAddresses, zone, domain, dns.TypeNS, 0,
 		)
 		if err != nil {
 			return nil, err
 		}
 
-		ansNS := extractNSSet(resp.Answer)
+		nsSet := nsSetFrom(resp, domain)
-		if len(ansNS) > 0 {
+		if len(nsSet) > 0 {
-			return ansNS, nil
+			return nsSet, nil
 		}
 
 		// An authoritative reply comes from the servers of the zone
 		// domain is in; it is not a referral, even when its authority
 		// section lists that zone's NS records. Without NS records in
-		// the answer, domain is not the zone's apex and has no
+		// the answer, domain has no nameservers of its own: it does
-		// nameservers of its own.
+		// not exist, when the reply is NXDOMAIN, or else it is not the
+		// zone's apex.
+		if resp.Authoritative && resp.Rcode == dns.RcodeNameError {
+			return nil, ErrNXDomain
+		}
+
 		if resp.Authoritative {
-			return nil, ErrNoNameservers
+			return []string{}, nil
 		}
 
 		authNS := extractNSSet(resp.Ns)
@@ -238,70 +256,249 @@ func (r *Resolver) followDelegation(
 			return r.resolveNSIterative(ctx, domain)
 		}
 
-		glue := extractGlue(resp.Extra)
+		servers, withoutAddresses = referralNameservers(resp)
-		nextServers := glueIPs(authNS, glue)
+		zone = referralZone(resp)
-
-		if len(nextServers) == 0 {
-			nextServers = r.resolveNSIPs(ctx, authNS)
-		}
-
-		if len(nextServers) == 0 {
-			return nil, ErrNoNameservers
-		}
-
-		servers = nextServers
 	}
 
 	return nil, ErrNoNameservers
 }
 
+// shuffled returns a copy of servers in the order shuffle puts them
+// in. The resolver passes rand.Shuffle, so each time it walks a list of
+// servers it starts at a random one, and no one server gets every
+// first query.
+func shuffled(
+	servers []string,
+	shuffle func(n int, swap func(i, j int)),
+) []string {
+	order := slices.Clone(servers)
+	shuffle(len(order), func(i, j int) {
+		order[i], order[j] = order[j], order[i]
+	})
+
+	return order
+}
+
+// queryServers asks servers, the servers of zone, about name in a random
+// order until one gives a usable reply. A server that times out, refuses
+// or gives a reply that is not usable is passed over for the next. When
+// every server refused, the error says so, and when they are the root
+// servers it is ErrIntercepted.
 func (r *Resolver) queryServers(
 	ctx context.Context,
 	servers []string,
+	zone string,
 	name string,
 	qtype uint16,
 ) (*dns.Msg, error) {
 	var lastErr error
 
-	for _, ip := range servers {
+	refused := 0
+
+	for _, ip := range shuffled(servers, rand.Shuffle) {
 		if checkCtx(ctx) != nil {
 			return nil, ErrContextCanceled
 		}
 
 		resp, err := r.queryDNS(ctx, ip, name, qtype)
+		if err == nil && !usableReply(resp, zone, name) {
+			err = fmt.Errorf(
+				"query %s @%s: %w", name, ip, ErrUnusableReply,
+			)
+		}
+
 		if err == nil {
 			return resp, nil
 		}
 
+		if errors.Is(err, ErrRefused) {
+			refused++
+		}
+
 		lastErr = err
 	}
 
+	if refused == len(servers) && zone == "." {
+		return nil, fmt.Errorf(
+			"every root server refused a query for %s: %w",
+			name, ErrIntercepted,
+		)
+	}
+
+	if refused == len(servers) {
+		return nil, fmt.Errorf(
+			"every server of %s refused a query for %s: %w",
+			zone, name, ErrRefused,
+		)
+	}
+
 	return nil, fmt.Errorf("all servers failed: %w", lastErr)
 }
 
+// isErrorReply reports whether msg is an error reply: one with any code
+// but NOERROR and NXDOMAIN, such as SERVFAIL, NOTIMP or FORMERR. An error
+// reply says nothing about the name's records.
+func isErrorReply(msg *dns.Msg) bool {
+	return msg.Rcode != dns.RcodeSuccess && msg.Rcode != dns.RcodeNameError
+}
+
+// usableReply reports whether resp, a reply from one of the servers of
+// zone to a query about name, is usable. An error reply such as SERVFAIL
+// is not. Nor is a referral, unless it refers the query to a zone below
+// zone that name is in: a server that refers it back to zone, up or
+// sideways does not serve zone as it should.
+func usableReply(resp *dns.Msg, zone string, name string) bool {
+	if isErrorReply(resp) {
+		return false
+	}
+
+	child := referralZone(resp)
+	if resp.Authoritative || len(resp.Answer) > 0 || child == "" {
+		return true
+	}
+
+	return child != zone && dns.IsSubDomain(zone, child) &&
+		dns.IsSubDomain(child, name)
+}
+
+// referralZone returns the zone a referral refers the query to: the
+// owner name of the NS records in resp's authority section, or "" when
+// there are none.
+func referralZone(resp *dns.Msg) string {
+	for _, rr := range resp.Ns {
+		if ns, ok := rr.(*dns.NS); ok {
+			return strings.ToLower(ns.Hdr.Name)
+		}
+	}
+
+	return ""
+}
+
+// nsSetFrom returns the NS set of domain that resp, a reply to a query
+// for domain's NS records, gives: the delegation in a referral to domain
+// itself, or else the NS records in the answer; empty when it gives
+// neither. A referral to domain comes from its parent zone's servers,
+// which all hold the same delegation, so the set does not depend on
+// which of them answered. domain's own servers, which can disagree about
+// their NS records, are then not asked.
+func nsSetFrom(resp *dns.Msg, domain string) []string {
+	if referralZone(resp) == domain {
+		return extractNSSet(resp.Ns)
+	}
+
+	return extractNSSet(resp.Answer)
+}
+
+// referralNameservers returns the IPv4 addresses that resp, a referral,
+// gives for the nameservers it names, and the names of the nameservers
+// it gives no address for.
+func referralNameservers(resp *dns.Msg) ([]string, []string) {
+	glue := extractGlue(resp.Extra)
+
+	var given, withoutAddresses []string
+
+	for _, ns := range extractNSSet(resp.Ns) {
+		ips := glueIPs([]string{ns}, glue)
+		if len(ips) == 0 {
+			withoutAddresses = append(withoutAddresses, ns)
+		}
+
+		given = append(given, ips...)
+	}
+
+	return given, withoutAddresses
+}
+
+// queryZone asks the servers of zone about name as queryServers does:
+// first those at given, the addresses a referral gave, and only when
+// none of them gives a usable reply, the nameservers named
+// withoutAddresses, once their addresses are looked up. depth is how
+// many lookups of a nameserver's address are under way, 0 in the walk
+// to a domain's nameservers; at maxLookupDepth, no address is looked
+// up. When the limit is why none was found, here or in a lookup this
+// one started, the error is ErrLookupDepthExceeded.
+func (r *Resolver) queryZone(
+	ctx context.Context,
+	given []string,
+	withoutAddresses []string,
+	zone string,
+	name string,
+	qtype uint16,
+	depth int,
+) (*dns.Msg, error) {
+	err := fmt.Errorf(
+		"no address for any nameserver of %s: %w", zone, ErrNoNameservers,
+	)
+
+	if len(given) > 0 {
+		var resp *dns.Msg
+
+		resp, err = r.queryServers(ctx, given, zone, name, qtype)
+		if err == nil {
+			return resp, nil
+		}
+	}
+
+	if len(withoutAddresses) == 0 {
+		return nil, err
+	}
+
+	if depth >= maxLookupDepth {
+		return nil, fmt.Errorf(
+			"addresses of the nameservers of %s not looked up: %w",
+			zone, ErrLookupDepthExceeded,
+		)
+	}
+
+	lookedUp, limitErr := r.resolveNSIPs(ctx, withoutAddresses, depth+1)
+	if limitErr != nil {
+		return nil, limitErr
+	}
+
+	if len(lookedUp) == 0 {
+		return nil, err
+	}
+
+	return r.queryServers(ctx, lookedUp, zone, name, qtype)
+}
+
+// resolveNSIPs returns the addresses of every nameserver in nsNames
+// whose name resolves, each looked up at depth (see resolveARecord).
+// The walk can then go on to the zone's other nameservers when one
+// gives no usable reply. When none resolves and the depth limit
+// stopped one of the lookups, it returns that lookup's error.
 func (r *Resolver) resolveNSIPs(
 	ctx context.Context,
 	nsNames []string,
-) []string {
+	depth int,
-	var ips []string
+) ([]string, error) {
+	var (
+		ips      []string
+		limitErr error
+	)
 
 	for _, ns := range nsNames {
-		resolved, err := r.resolveARecord(ctx, ns)
+		resolved, err := r.resolveARecord(ctx, ns, depth)
-		if err == nil {
+
+		switch {
+		case err == nil:
 			ips = append(ips, resolved...)
+		case errors.Is(err, ErrLookupDepthExceeded):
+			limitErr = err
+		}
 	}
 
 	if len(ips) > 0 {
-			break
+		return ips, nil
-		}
 	}
 
-	return ips
+	return nil, limitErr
 }
 
 // resolveNSIterative queries for NS records using iterative
 // resolution as a fallback when followDelegation finds no
-// authoritative answer in the delegation chain.
+// authoritative answer in the delegation chain. Its result means what
+// followDelegation's does.
 func (r *Resolver) resolveNSIterative(
 	ctx context.Context,
 	domain string,
@@ -312,6 +509,7 @@ func (r *Resolver) resolveNSIterative(
 
 	domain = dns.Fqdn(domain)
 	servers := rootServerList()
+	zone := "."
 
 	for range maxDelegation {
 		if checkCtx(ctx) != nil {
@@ -319,17 +517,27 @@ func (r *Resolver) resolveNSIterative(
 		}
 
 		resp, err := r.queryServers(
-			ctx, servers, domain, dns.TypeNS,
+			ctx, servers, zone, domain, dns.TypeNS,
 		)
 		if err != nil {
 			return nil, err
 		}
 
-		nsNames := extractNSSet(resp.Answer)
+		nsNames := nsSetFrom(resp, domain)
 		if len(nsNames) > 0 {
 			return nsNames, nil
 		}
 
+		// As in followDelegation: domain has no nameservers of its
+		// own.
+		if resp.Authoritative && resp.Rcode == dns.RcodeNameError {
+			return nil, ErrNXDomain
+		}
+
+		if resp.Authoritative {
+			return []string{}, nil
+		}
+
 		// Follow delegation.
 		authNS := extractNSSet(resp.Ns)
 		if len(authNS) == 0 {
@@ -344,16 +552,20 @@ func (r *Resolver) resolveNSIterative(
 		}
 
 		servers = nextServers
+		zone = referralZone(resp)
 	}
 
 	return nil, ErrNoNameservers
 }
 
-// resolveARecord resolves a hostname to IPv4 addresses using
+// resolveARecord resolves a hostname, a nameserver's name, to IPv4
-// iterative resolution through the delegation chain.
+// addresses using iterative resolution through the delegation chain.
+// depth is how many lookups of a nameserver's address are under way,
+// this one included: 1 for a lookup that no other lookup started.
 func (r *Resolver) resolveARecord(
 	ctx context.Context,
 	hostname string,
+	depth int,
 ) ([]string, error) {
 	if checkCtx(ctx) != nil {
 		return nil, ErrContextCanceled
@@ -361,14 +573,18 @@ func (r *Resolver) resolveARecord(
 
 	hostname = dns.Fqdn(hostname)
 	servers := rootServerList()
+	zone := "."
+
+	var withoutAddresses []string
 
 	for range maxDelegation {
 		if checkCtx(ctx) != nil {
 			return nil, ErrContextCanceled
 		}
 
-		resp, err := r.queryServers(
+		resp, err := r.queryZone(
-			ctx, servers, hostname, dns.TypeA,
+			ctx, servers, withoutAddresses, zone, hostname, dns.TypeA,
+			depth,
 		)
 		if err != nil {
 			return nil, fmt.Errorf(
@@ -395,17 +611,8 @@ func (r *Resolver) resolveARecord(
 			break
 		}
 
-		glue := extractGlue(resp.Extra)
+		servers, withoutAddresses = referralNameservers(resp)
-		nextServers := glueIPs(authNS, glue)
+		zone = referralZone(resp)
-
-		if len(nextServers) == 0 {
-			// Resolve NS IPs iteratively — but guard
-			// against infinite recursion by using only
-			// already-resolved servers.
-			break
-		}
-
-		servers = nextServers
 	}
 
 	return nil, fmt.Errorf(
@@ -415,12 +622,24 @@ func (r *Resolver) resolveARecord(
 
 // FindAuthoritativeNameservers traces the delegation chain from
 // root servers to discover all authoritative nameservers for the
-// given domain. For a name that is not a zone apex it tries each
+// given domain, as the delegation from its parent zone's servers lists
-// parent name in turn, so it returns the nameservers of the zone the
+// them. When the servers asked answer that the name has no delegation
-// name is in.
+// of its own, or does not exist, it tries each parent name in turn, so
+// it returns the nameservers of the zone the name is in. When they do
+// not answer, it returns the error and tries no parent name.
 func (r *Resolver) FindAuthoritativeNameservers(
 	ctx context.Context,
 	domain string,
+) ([]string, error) {
+	return r.findAuthoritativeNameservers(ctx, domain, rootServerList())
+}
+
+// findAuthoritativeNameservers is FindAuthoritativeNameservers with each
+// walk starting at servers, the root servers.
+func (r *Resolver) findAuthoritativeNameservers(
+	ctx context.Context,
+	domain string,
+	servers []string,
 ) ([]string, error) {
 	if checkCtx(ctx) != nil {
 		return nil, ErrContextCanceled
@@ -436,10 +655,12 @@ func (r *Resolver) FindAuthoritativeNameservers(
 
 		candidate := strings.Join(labels[i:], ".") + "."
 
-		nsNames, err := r.followDelegation(
+		nsNames, err := r.followDelegation(ctx, candidate, servers)
-			ctx, candidate, rootServerList(),
+		if err != nil && !errors.Is(err, ErrNXDomain) {
-		)
+			return nil, err
-		if err == nil && len(nsNames) > 0 {
+		}
+
+		if len(nsNames) > 0 {
 			sort.Strings(nsNames)
 
 			return nsNames, nil
@@ -449,25 +670,52 @@ func (r *Resolver) FindAuthoritativeNameservers(
 	return nil, ErrNoNameservers
 }
 
+// recordTypes returns the record types a nameserver is asked for when a
+// name is checked.
+func recordTypes() []uint16 {
+	return []uint16{
+		dns.TypeA, dns.TypeAAAA, dns.TypeCNAME,
+		dns.TypeMX, dns.TypeTXT, dns.TypeSRV,
+		dns.TypeCAA, dns.TypeNS,
+	}
+}
+
+// addressTypes returns the record types ResolveIPAddresses asks for,
+// the only ones it reads.
+func addressTypes() []uint16 {
+	return []uint16{dns.TypeA, dns.TypeAAAA, dns.TypeCNAME}
+}
+
 // QueryNameserver queries a specific nameserver for all record
 // types and builds a NameserverResponse.
 func (r *Resolver) QueryNameserver(
 	ctx context.Context,
 	nsHostname string,
 	hostname string,
+) (*NameserverResponse, error) {
+	return r.queryNameserver(ctx, nsHostname, hostname, recordTypes())
+}
+
+// queryNameserver queries a specific nameserver for the record types
+// in qtypes and builds a NameserverResponse.
+func (r *Resolver) queryNameserver(
+	ctx context.Context,
+	nsHostname string,
+	hostname string,
+	qtypes []uint16,
 ) (*NameserverResponse, error) {
 	if checkCtx(ctx) != nil {
 		return nil, ErrContextCanceled
 	}
 
-	nsIPs, err := r.resolveARecord(ctx, nsHostname)
+	nsIPs, err := r.resolveARecord(ctx, nsHostname, 1)
 	if err != nil {
 		return nil, fmt.Errorf("resolving NS %s: %w", nsHostname, err)
 	}
 
 	hostname = dns.Fqdn(hostname)
 
-	return r.queryAllTypes(ctx, nsHostname, nsIPs[0], hostname)
+	return r.queryTypes(ctx, nsHostname, nsIPs[0], hostname, qtypes)
 }
 
 // QueryNameserverIP queries a nameserver by its IP address directly,
@@ -484,14 +732,15 @@ func (r *Resolver) QueryNameserverIP(
 
 	hostname = dns.Fqdn(hostname)
 
-	return r.queryAllTypes(ctx, nsHostname, nsIP, hostname)
+	return r.queryTypes(ctx, nsHostname, nsIP, hostname, recordTypes())
 }
 
-func (r *Resolver) queryAllTypes(
+func (r *Resolver) queryTypes(
 	ctx context.Context,
 	nsHostname string,
 	nsIP string,
 	hostname string,
+	qtypes []uint16,
 ) (*NameserverResponse, error) {
 	resp := &NameserverResponse{
 		Nameserver: nsHostname,
@@ -499,12 +748,6 @@ func (r *Resolver) queryAllTypes(
 		Status:     StatusOK,
 	}
 
-	qtypes := []uint16{
-		dns.TypeA, dns.TypeAAAA, dns.TypeCNAME,
-		dns.TypeMX, dns.TypeTXT, dns.TypeSRV,
-		dns.TypeCAA, dns.TypeNS,
-	}
-
 	state := r.queryEachType(ctx, nsIP, hostname, qtypes, resp)
 	classifyResponse(resp, state)
 
@@ -513,13 +756,20 @@ func (r *Resolver) queryAllTypes(
 
 type queryState struct {
 	gotNXDomain   bool
-	gotSERVFAIL bool
+	gotErrorReply bool
+	errorReply    string // its code, such as SERVFAIL, or number if unnamed
 	gotRefused    bool
 	gotTimeout    bool
+	gotReferral   bool
 	netErr        error
 	hasRecords    bool
+	answered      bool
 }
 
+// queryEachType asks the nameserver at nsIP about hostname once for each
+// record type in qtypes, and lists in resp.FailedTypes the types whose
+// query got no usable reply, logging each with the reason unless ctx was
+// cancelled: shutdown cancels it, and a query it cut short did not fail.
 func (r *Resolver) queryEachType(
 	ctx context.Context,
 	nsIP string,
@@ -534,7 +784,34 @@ func (r *Resolver) queryEachType(
 			break
 		}
 
-		r.querySingleType(ctx, nsIP, hostname, qtype, resp, &state)
+		err := r.querySingleType(ctx, nsIP, hostname, qtype, resp, &state)
+		if err == nil {
+			state.answered = true
+
+			continue
+		}
+
+		rtype := dns.TypeToString[qtype]
+		resp.FailedTypes = append(resp.FailedTypes, rtype)
+
+		if errors.Is(ctx.Err(), context.Canceled) {
+			continue
+		}
+
+		r.log.Warn(
+			"record type query failed",
+			"hostname", hostname,
+			"nameserver", resp.Nameserver,
+			"type", rtype,
+			"error", err,
+		)
+	}
+
+	// The reply about another type can carry the name's CNAME. When the
+	// query for CNAME itself failed, that is left out too, so Records
+	// holds nothing for a failed type.
+	for _, rtype := range resp.FailedTypes {
+		delete(resp.Records, rtype)
 	}
 
 	for k := range resp.Records {
@@ -544,6 +821,9 @@ func (r *Resolver) queryEachType(
 	return state
 }
 
+// querySingleType asks the nameserver at nsIP about hostname's records
+// of type qtype. It returns nil when the nameserver answered: with
+// records, with none, or with NXDOMAIN; otherwise it returns why not.
 func (r *Resolver) querySingleType(
 	ctx context.Context,
 	nsIP string,
@@ -551,7 +831,7 @@ func (r *Resolver) querySingleType(
 	qtype uint16,
 	resp *NameserverResponse,
 	state *queryState,
-) {
+) error {
 	msg, err := r.queryDNS(ctx, nsIP, hostname, qtype)
 	if err != nil {
 		switch {
@@ -563,24 +843,70 @@ func (r *Resolver) querySingleType(
 			state.netErr = err
 		}
 
-		return
+		return err
 	}
 
+	return readReply(msg, resp, state)
+}
+
+// readReply adds to resp the records in msg, a nameserver's reply to a
+// query about one record type. It returns nil when the nameserver
+// answered: with records, with none, or with NXDOMAIN; otherwise it
+// returns why not.
+func readReply(
+	msg *dns.Msg,
+	resp *NameserverResponse,
+	state *queryState,
+) error {
 	if msg.Rcode == dns.RcodeNameError {
 		state.gotNXDomain = true
 
-		return
+		return nil
 	}
 
-	if msg.Rcode == dns.RcodeServerFailure {
+	if isErrorReply(msg) {
-		state.gotSERVFAIL = true
+		state.gotErrorReply = true
 
-		return
+		code, named := dns.RcodeToString[msg.Rcode]
+		if !named {
+			code = strconv.Itoa(msg.Rcode)
+		}
+
+		state.errorReply = code
+
+		return fmt.Errorf(
+			"server returned %s: %w", state.errorReply, ErrUnusableReply,
+		)
+	}
+
+	// A reply with no answer that lists other nameservers, from a server
+	// that does not hold the name's zone, is a referral and says nothing
+	// about the name's records. A server named in the delegation that
+	// does not hold the zone may send one.
+	if !msg.Authoritative && len(msg.Answer) == 0 &&
+		len(extractNSSet(msg.Ns)) > 0 {
+		state.gotReferral = true
+
+		return fmt.Errorf("server returned a referral: %w", ErrUnusableReply)
+	}
+
+	// A reply still truncated is one whose TCP retry failed, and holds
+	// only the records that fit.
+	if msg.Truncated {
+		state.netErr = ErrTruncated
+
+		return ErrTruncated
 	}
 
 	collectAnswerRecords(msg, resp, state)
+
+	return nil
 }
 
+// collectAnswerRecords adds the records in msg's answer to resp, each
+// value once per record type. For a name with a CNAME, a nameserver
+// answers a query of any type with that CNAME, so the same value comes
+// in the answer to every type asked for.
 func collectAnswerRecords(
 	msg *dns.Msg,
 	resp *NameserverResponse,
@@ -593,9 +919,12 @@ func collectAnswerRecords(
 		}
 
 		typeName := dns.TypeToString[rr.Header().Rrtype]
+		if !slices.Contains(resp.Records[typeName], val) {
 			resp.Records[typeName] = append(
 				resp.Records[typeName], val,
 			)
+		}
+
 		state.hasRecords = true
 	}
 }
@@ -610,23 +939,32 @@ func isTimeout(err error) bool {
 	return false
 }
 
+// classifyResponse sets the nameserver's status. One that answered no
+// record type has failed, and Error says why; one that answered some has
+// the status of those answers. It has no data only when every type
+// answered with no records: a type in FailedTypes may have records, so a
+// nameserver with one stays ok.
 func classifyResponse(resp *NameserverResponse, state queryState) {
 	switch {
 	case state.gotNXDomain && !state.hasRecords:
 		resp.Status = StatusNXDomain
-	case state.gotTimeout && !state.hasRecords:
+	case state.gotTimeout && !state.answered:
 		resp.Status = StatusTimeout
 		resp.Error = "all queries timed out"
-	case state.gotSERVFAIL && !state.hasRecords:
+	case state.gotErrorReply && !state.answered:
 		resp.Status = StatusError
-		resp.Error = "server returned SERVFAIL"
+		resp.Error = "server returned " + state.errorReply
-	case state.gotRefused && !state.hasRecords:
+	case state.gotRefused && !state.answered:
 		resp.Status = StatusError
 		resp.Error = "server returned REFUSED"
-	case state.netErr != nil && !state.hasRecords:
+	case state.netErr != nil && !state.answered:
 		resp.Status = StatusError
 		resp.Error = "network error: " + state.netErr.Error()
-	case !state.hasRecords && !state.gotNXDomain:
+	case state.gotReferral && !state.answered:
+		resp.Status = StatusError
+		resp.Error = "server returned a referral"
+	// An NXDOMAIN reply with no records was taken by the first case.
+	case !state.hasRecords && len(resp.FailedTypes) == 0:
 		resp.Status = StatusNoData
 	}
 }
@@ -678,18 +1016,19 @@ func (r *Resolver) QueryAllNameservers(
 		return nil, err
 	}
 
-	return r.queryEachNS(ctx, nameservers, hostname)
+	return r.queryEachNS(ctx, nameservers, hostname, recordTypes())
 }
 
 func (r *Resolver) queryEachNS(
 	ctx context.Context,
 	nameservers []string,
 	hostname string,
+	qtypes []uint16,
 ) (map[string]*NameserverResponse, error) {
 	results := make(map[string]*NameserverResponse)
 
 	for _, ns := range nameservers {
-		resp, err := r.QueryNameserver(ctx, ns, hostname)
+		resp, err := r.queryNameserver(ctx, ns, hostname, qtypes)
 
 		// A query the context cut short says nothing about the
 		// nameserver, so it must not be returned as its failure.
@@ -714,12 +1053,22 @@ func (r *Resolver) queryEachNS(
 	return results, nil
 }
 
-// LookupNS returns the NS record set for a domain.
+// LookupNS returns the NS record set of a domain, as the delegation from
+// its parent zone's servers lists it, and never a parent name's. When
+// they answer that the domain does not exist, the error is ErrNXDomain.
+// When they answer that it has no delegation of its own, the set is
+// empty and there is no error.
 func (r *Resolver) LookupNS(
 	ctx context.Context,
 	domain string,
 ) ([]string, error) {
-	return r.FindAuthoritativeNameservers(ctx, domain)
+	if checkCtx(ctx) != nil {
+		return nil, ErrContextCanceled
+	}
+
+	return r.followDelegation(
+		ctx, dns.Fqdn(strings.ToLower(domain)), rootServerList(),
+	)
 }
 
 // LookupAllRecords performs iterative resolution to find all DNS
@@ -734,7 +1083,10 @@ func (r *Resolver) LookupAllRecords(
 }
 
 // ResolveIPAddresses resolves a hostname to all IPv4 and IPv6
-// addresses, following CNAME chains up to MaxCNAMEDepth.
+// addresses, following CNAME chains up to MaxCNAMEDepth. It asks each
+// nameserver of the name's zone for its A, AAAA and CNAME records only.
+// When no nameserver of the name's zone answered, it returns an error
+// rather than no addresses.
 func (r *Resolver) ResolveIPAddresses(
 	ctx context.Context,
 	hostname string,
@@ -755,12 +1107,20 @@ func (r *Resolver) resolveIPWithCNAME(
 		return nil, ErrCNAMEDepthExceeded
 	}
 
-	results, err := r.QueryAllNameservers(ctx, hostname)
+	nameservers, err := r.FindAuthoritativeNameservers(ctx, hostname)
 	if err != nil {
 		return nil, err
 	}
 
-	ips, cnameTarget := collectIPs(results)
+	results, err := r.queryEachNS(ctx, nameservers, hostname, addressTypes())
+	if err != nil {
+		return nil, err
+	}
+
+	ips, cnameTarget, err := collectIPs(results)
+	if err != nil {
+		return nil, fmt.Errorf("resolving %s: %w", hostname, err)
+	}
 
 	if len(ips) == 0 && cnameTarget != "" {
 		return r.resolveIPWithCNAME(ctx, cnameTarget, depth+1)
@@ -771,16 +1131,31 @@ func (r *Resolver) resolveIPWithCNAME(
 	return ips, nil
 }
 
+// collectIPs returns the addresses in the nameservers' answers and the
+// first CNAME target among them. A nameserver whose query for one of the
+// types failed gave only part of the addresses, and is left out. It
+// returns ErrNoNameserverAnswered when every nameserver timed out,
+// failed, returned a referral or was left out: that is not a name with
+// no addresses.
 func collectIPs(
 	results map[string]*NameserverResponse,
-) ([]string, string) {
+) ([]string, string, error) {
 	seen := make(map[string]bool)
 
 	var ips []string
 
 	var cnameTarget string
 
+	answered := false
+
 	for _, resp := range results {
+		if resp.Status == StatusTimeout || resp.Status == StatusError ||
+			len(resp.FailedTypes) > 0 {
+			continue
+		}
+
+		answered = true
+
 		if resp.Status == StatusNXDomain {
 			continue
 		}
@@ -804,5 +1179,9 @@ func collectIPs(
 		}
 	}
 
-	return ips, cnameTarget
+	if !answered {
+		return nil, "", ErrNoNameserverAnswered
+	}
+
+	return ips, cnameTarget, nil
 }

internal/resolver/iterative_internal_test.go

@@ -0,0 +1,149 @@
+package resolver
+
+import (
+	"strconv"
+	"syscall"
+	"testing"
+
+	"github.com/miekg/dns"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// TestClassifyResponse sets a nameserver's status from the results of
+// its queries and the record types whose query failed, built here. One
+// that answered some record types, even with no records, has not failed
+// when its query for another type got no usable reply, whatever the
+// reason, and is ok, not nodata: that type may have records. One whose
+// every query got none has failed. Only one whose every type answered
+// with no records is nodata.
+func TestClassifyResponse(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name        string
+		results     queryState
+		failedTypes []string
+		wantStatus  string
+		wantError   string
+	}{
+		{
+			"every type answered with no records",
+			queryState{answered: true},
+			nil,
+			StatusNoData, "",
+		},
+		{
+			"some types answered with no records, another timed out",
+			queryState{answered: true, gotTimeout: true},
+			[]string{"A"},
+			StatusOK, "",
+		},
+		{
+			"some types answered with no records, another got SERVFAIL",
+			queryState{
+				answered: true, gotErrorReply: true, errorReply: "SERVFAIL",
+			},
+			[]string{"A"},
+			StatusOK, "",
+		},
+		{
+			"some types answered with no records, another was refused",
+			queryState{answered: true, gotRefused: true},
+			[]string{"A"},
+			StatusOK, "",
+		},
+		{
+			"some types answered with no records, another got a network error",
+			queryState{answered: true, netErr: syscall.ECONNREFUSED},
+			[]string{"A"},
+			StatusOK, "",
+		},
+		{
+			"some types answered with no records, another's reply was " +
+				"truncated and its retry over TCP failed",
+			queryState{answered: true, netErr: ErrTruncated},
+			[]string{"TXT"},
+			StatusOK, "",
+		},
+		{
+			"some types answered with no records, another got a referral",
+			queryState{answered: true, gotReferral: true},
+			[]string{"A"},
+			StatusOK, "",
+		},
+		{
+			"every query timed out",
+			queryState{gotTimeout: true},
+			[]string{"A", "AAAA", "CNAME"},
+			StatusTimeout, "all queries timed out",
+		},
+		{
+			"every query got NOTIMP",
+			queryState{gotErrorReply: true, errorReply: "NOTIMP"},
+			[]string{"A", "AAAA", "CNAME"},
+			StatusError, "server returned NOTIMP",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			resp := &NameserverResponse{Status: StatusOK, FailedTypes: tt.failedTypes}
+			classifyResponse(resp, tt.results)
+
+			assert.Equal(t, tt.wantStatus, resp.Status)
+			assert.Equal(t, tt.wantError, resp.Error)
+			assert.Equal(t, tt.failedTypes, resp.FailedTypes)
+		})
+	}
+}
+
+// TestReadReply checks which replies to a query about one record type,
+// built here, are an answer: one with the code NOERROR or NXDOMAIN. A
+// reply with any other code is not, and the type's query has failed; a
+// nameserver whose only reply it is has failed, and Error gives the
+// code, or its number when the code has no name.
+func TestReadReply(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		rcode      int
+		wantStatus string
+		wantError  string
+	}{
+		{dns.RcodeSuccess, StatusNoData, ""},
+		{dns.RcodeNameError, StatusNXDomain, ""},
+		{dns.RcodeServerFailure, StatusError, "server returned SERVFAIL"},
+		{dns.RcodeNotImplemented, StatusError, "server returned NOTIMP"},
+		{dns.RcodeFormatError, StatusError, "server returned FORMERR"},
+		{12, StatusError, "server returned 12"}, // unassigned, no name
+	}
+
+	for _, tt := range tests {
+		t.Run(strconv.Itoa(tt.rcode), func(t *testing.T) {
+			t.Parallel()
+
+			msg := new(dns.Msg)
+			msg.Authoritative = true
+			msg.Rcode = tt.rcode
+
+			resp := &NameserverResponse{Records: map[string][]string{}}
+
+			var state queryState
+
+			err := readReply(msg, resp, &state)
+			classifyResponse(resp, state)
+
+			if tt.wantStatus == StatusError {
+				require.ErrorIs(t, err, ErrUnusableReply)
+			} else {
+				require.NoError(t, err)
+			}
+
+			assert.Equal(t, tt.wantStatus, resp.Status)
+			assert.Equal(t, tt.wantError, resp.Error)
+		})
+	}
+}

internal/resolver/iterative_test.go

@@ -1,14 +1,210 @@
 package resolver_test
 
 import (
+	"math/rand/v2"
+	"slices"
 	"testing"
 
 	"github.com/miekg/dns"
 	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
 
 	"sneak.berlin/go/dnswatcher/internal/resolver"
 )
 
+// TestCollectIPs_OneAnswerIsEnough checks that one nameserver answering
+// NXDOMAIN says the name has no addresses, though the other timed out.
+func TestCollectIPs_OneAnswerIsEnough(t *testing.T) {
+	t.Parallel()
+
+	ips, _, err := resolver.CollectIPs(
+		map[string]*resolver.NameserverResponse{
+			"ns1.example.": {Status: resolver.StatusTimeout},
+			"ns2.example.": {Status: resolver.StatusNXDomain},
+		},
+	)
+	require.NoError(t, err)
+	assert.Empty(t, ips)
+}
+
+// TestCollectIPs_FailedIsNoAnswer checks that nameservers that all have
+// status error, from a refusal, a server failure, a network error or a
+// referral, are no answer rather than a name with no addresses.
+func TestCollectIPs_FailedIsNoAnswer(t *testing.T) {
+	t.Parallel()
+
+	ips, _, err := resolver.CollectIPs(
+		map[string]*resolver.NameserverResponse{
+			"ns1.example.": {Status: resolver.StatusError},
+			"ns2.example.": {Status: resolver.StatusError},
+		},
+	)
+	require.ErrorIs(t, err, resolver.ErrNoNameserverAnswered)
+	assert.Empty(t, ips)
+}
+
+// TestCollectIPs_FailedTypeIsNoAnswer checks that a nameserver whose
+// query for one of the types failed is no answer: its addresses are
+// only part of them.
+func TestCollectIPs_FailedTypeIsNoAnswer(t *testing.T) {
+	t.Parallel()
+
+	ips, _, err := resolver.CollectIPs(
+		map[string]*resolver.NameserverResponse{
+			nsExample1: {
+				Records:     map[string][]string{"A": {"192.0.2.1"}},
+				FailedTypes: []string{"AAAA"},
+				Status:      resolver.StatusOK,
+			},
+		},
+	)
+	require.ErrorIs(t, err, resolver.ErrNoNameserverAnswered)
+	assert.Empty(t, ips)
+}
+
+const (
+	// exampleCom is the zone most cases of TestUsableReply and
+	// TestNSSetFrom are about, and wwwExampleCom a name in it.
+	exampleCom    = "example.com."
+	wwwExampleCom = "www.example.com."
+
+	// exampleNS is the server the NS records nsRecord builds name.
+	exampleNS = "ns1.example.net."
+)
+
+// nsRecord builds an NS record that names a server of zone.
+func nsRecord(zone string) *dns.NS {
+	return &dns.NS{
+		Hdr: dns.RR_Header{
+			Name: zone, Rrtype: dns.TypeNS, Class: dns.ClassINET,
+		},
+		Ns: exampleNS,
+	}
+}
+
+// referralTo builds a reply that refers the query to the servers of
+// zone.
+func referralTo(zone string) *dns.Msg {
+	msg := new(dns.Msg)
+	msg.Ns = []dns.RR{nsRecord(zone)}
+
+	return msg
+}
+
+// TestUsableReply checks which replies from one of a zone's servers are
+// used. A reply that is not usable moves the query on to the zone's
+// next server.
+func TestUsableReply(t *testing.T) {
+	t.Parallel()
+
+	servfail := new(dns.Msg)
+	servfail.Rcode = dns.RcodeServerFailure
+
+	answer := new(dns.Msg)
+	answer.Authoritative = true
+	answer.Answer = []dns.RR{nsRecord(exampleCom)}
+
+	nxdomain := new(dns.Msg)
+	nxdomain.Authoritative = true
+	nxdomain.Rcode = dns.RcodeNameError
+
+	tests := []struct {
+		name  string
+		resp  *dns.Msg
+		zone  string
+		query string
+		want  bool
+	}{
+		{
+			name: "SERVFAIL", resp: servfail,
+			zone: exampleCom, query: exampleCom, want: false,
+		},
+		{
+			name: "answer", resp: answer,
+			zone: exampleCom, query: exampleCom, want: true,
+		},
+		{
+			name: "NXDOMAIN", resp: nxdomain,
+			zone: ".", query: exampleCom, want: true,
+		},
+		{
+			name: "root refers to com", resp: referralTo("com."),
+			zone: ".", query: exampleCom, want: true,
+		},
+		{
+			name: "com refers to example.com", resp: referralTo(exampleCom),
+			zone: "com.", query: wwwExampleCom, want: true,
+		},
+		{
+			name: "referral back to the zone", resp: referralTo(exampleCom),
+			zone: exampleCom, query: exampleCom, want: false,
+		},
+		{
+			name: "referral up to the root", resp: referralTo("."),
+			zone: exampleCom, query: exampleCom, want: false,
+		},
+		{
+			name: "referral sideways", resp: referralTo("net."),
+			zone: ".", query: exampleCom, want: false,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			assert.Equal(t, tt.want,
+				resolver.UsableReply(tt.resp, tt.zone, tt.query),
+			)
+		})
+	}
+}
+
+// TestNSSetFrom checks which NS set a reply gives for a domain; a set
+// that is not empty ends the walk. The referral to example.com that
+// com's servers all send alike gives its delegation, so the set is the
+// same whichever of them answered, and example.com's own servers, which
+// can disagree, are not asked.
+func TestNSSetFrom(t *testing.T) {
+	t.Parallel()
+
+	answer := new(dns.Msg)
+	answer.Authoritative = true
+	answer.Answer = []dns.RR{nsRecord(exampleCom)}
+
+	tests := []struct {
+		name   string
+		resp   *dns.Msg
+		domain string
+		want   []string
+	}{
+		{
+			name: "com refers to example.com", resp: referralTo(exampleCom),
+			domain: exampleCom, want: []string{exampleNS},
+		},
+		{
+			name: "com refers on, for www.example.com",
+			resp: referralTo(exampleCom), domain: wwwExampleCom,
+			want: nil,
+		},
+		{
+			name: "answer from a server that holds example.com",
+			resp: answer, domain: exampleCom,
+			want: []string{exampleNS},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			assert.ElementsMatch(t, tt.want,
+				resolver.NSSetFrom(tt.resp, tt.domain),
+			)
+		})
+	}
+}
+
 func TestExtractRecordValue_LetterCase(t *testing.T) {
 	t.Parallel()
 
@@ -60,3 +256,62 @@ func TestExtractRecordValue_LetterCase(t *testing.T) {
 		})
 	}
 }
+
+// TestCollectAnswerRecords_CNAMEOnce collects the answers a nameserver
+// gives for a name with a CNAME, one for each record type a check asks
+// for. Each answer holds the CNAME, which must be stored once.
+func TestCollectAnswerRecords_CNAMEOnce(t *testing.T) {
+	t.Parallel()
+
+	cname := &dns.CNAME{
+		Hdr: dns.RR_Header{
+			Name: "git.eeqj.de.", Rrtype: dns.TypeCNAME, Class: dns.ClassINET,
+		},
+		Target: "fsn1app1.datavi.be.",
+	}
+
+	resp := &resolver.NameserverResponse{Records: map[string][]string{}}
+
+	for _, qtype := range []uint16{
+		dns.TypeA, dns.TypeAAAA, dns.TypeCNAME, dns.TypeMX,
+		dns.TypeTXT, dns.TypeSRV, dns.TypeCAA, dns.TypeNS,
+	} {
+		msg := new(dns.Msg)
+		msg.SetQuestion("git.eeqj.de.", qtype)
+		msg.Answer = []dns.RR{cname}
+
+		resolver.CollectAnswerRecords(msg, resp)
+	}
+
+	assert.Equal(t,
+		map[string][]string{"CNAME": {"fsn1app1.datavi.be."}},
+		resp.Records,
+	)
+}
+
+// TestShuffled shuffles the root servers with many seeds. Every order
+// must hold each root server once, so each is tried before a
+// resolution fails; each root server must come first for some seed, so
+// no one root server gets every first query; and the list passed in
+// must be left as it was.
+func TestShuffled(t *testing.T) {
+	t.Parallel()
+
+	const seeds = 1000
+
+	roots := resolver.RootServerList()
+	before := slices.Clone(roots)
+	first := make(map[string]bool)
+
+	for seed := range uint64(seeds) {
+		rng := rand.New(rand.NewPCG(seed, 0)) //nolint:gosec // seeded on purpose
+		order := resolver.Shuffled(roots, rng.Shuffle)
+
+		assert.ElementsMatch(t, roots, order)
+
+		first[order[0]] = true
+	}
+
+	assert.Len(t, first, len(roots))
+	assert.Equal(t, before, roots)
+}

internal/resolver/livedns_test.go

@@ -187,8 +187,8 @@ func liveFindAuthoritative(
 	return out
 }
 
-// liveLookupNS is liveFindAuthoritative through the LookupNS entry
+// liveLookupNS looks up the NS record set of domain, a domain that has
-// point, so that both entry points stay independently exercised.
+// one, retrying until the delegation chain can be walked.
 func liveLookupNS(
 	t *testing.T,
 	r *resolver.Resolver,
@@ -226,11 +226,17 @@ func liveLookupNS(
 // liveQueryNameserver queries one nameserver, retrying while that
 // nameserver fails to answer. NXDOMAIN and NODATA are answers and
 // are returned to the caller to assert on.
+//
+// QueryNameserver sends one query per record type, so one lost query
+// leaves its type out of an answer that is otherwise fine. A test names
+// in types the record types it reads; an answer holding records of none
+// of them is retried too.
 func liveQueryNameserver(
 	t *testing.T,
 	r *resolver.Resolver,
 	nameserver string,
 	hostname string,
+	types ...string,
 ) *resolver.NameserverResponse {
 	t.Helper()
 
@@ -260,6 +266,18 @@ func liveQueryNameserver(
 				)
 			}
 
+			hasRecords := func(recordType string) bool {
+				return len(resp.Records[recordType]) > 0
+			}
+
+			if len(types) > 0 && !slices.ContainsFunc(types, hasRecords) {
+				return fmt.Errorf(
+					"%w: %s returned no %s records",
+					livednstest.ErrNoAnswer, nameserver,
+					strings.Join(types, " or "),
+				)
+			}
+
 			out = resp
 
 			return nil
@@ -383,3 +401,37 @@ func liveResolveIPsAllowingEmpty(
 
 	return out
 }
+
+// liveResolveNSIPs looks up the addresses of the nameservers named
+// names, retrying until there are at least atLeast of them: a name
+// whose lookup got no reply is left out of the result, not an error.
+func liveResolveNSIPs(
+	t *testing.T,
+	r *resolver.Resolver,
+	names []string,
+	atLeast int,
+) []string {
+	t.Helper()
+
+	var out []string
+
+	livednstest.Retry(
+		t,
+		"ResolveNSIPs("+strings.Join(names, ", ")+")",
+		func(ctx context.Context) error {
+			ips := r.ResolveNSIPs(ctx, names)
+			if len(ips) < atLeast {
+				return fmt.Errorf(
+					"%w: %d addresses, expected at least %d",
+					livednstest.ErrNoAnswer, len(ips), atLeast,
+				)
+			}
+
+			out = ips
+
+			return nil
+		},
+	)
+
+	return out
+}

internal/resolver/resolver.go

@@ -31,9 +31,13 @@ type Params struct {
 }
 
 // NameserverResponse holds one nameserver's response for a query.
+// FailedTypes lists the record types whose query got no usable reply,
+// and Records holds nothing for them: their records are not known. When
+// no record type got one, Status and Error say the nameserver failed.
 type NameserverResponse struct {
 	Nameserver  string
 	Records     map[string][]string
+	FailedTypes []string
 	Status      string
 	Error       string
 }

internal/resolver/resolver_test.go

@@ -1,7 +1,9 @@
 package resolver_test
 
 import (
+	"bytes"
 	"context"
+	"errors"
 	"fmt"
 	"log/slog"
 	"net"
@@ -11,6 +13,7 @@ import (
 	"testing"
 	"time"
 
+	"github.com/miekg/dns"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 
@@ -22,6 +25,13 @@ import (
 // Test helpers
 // ----------------------------------------------------------------
 
+// nonexistentDomain is a .com domain that does not exist.
+const nonexistentDomain = "dnswatcher-test-does-not-exist.com"
+
+// noAnswerAddress is 192.0.2.1, a documentation address: nothing
+// answers there.
+const noAnswerAddress = "192.0.2.1"
+
 func newTestResolver(t *testing.T) *resolver.Resolver {
 	t.Helper()
 
@@ -85,6 +95,47 @@ func TestFindAuthoritativeNameservers_Subdomain(
 	assert.Equal(t, fromZone, fromHost)
 }
 
+// TestFindAuthoritativeNameservers_DelegatedSubdomain looks up the
+// nameservers of www.cs.cmu.edu, a name in cs.cmu.edu, a zone that
+// cmu.edu delegates to other servers. The servers of cs.cmu.edu answer
+// that the name has no delegation of its own, so it gets their names,
+// not those of the cmu.edu servers. Every referral on the way gives the
+// nameservers' addresses, so the walk sends few queries.
+func TestFindAuthoritativeNameservers_DelegatedSubdomain(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	r := newTestResolver(t)
+	fromHost := liveFindAuthoritative(t, r, "www.cs.cmu.edu")
+	fromZone := liveLookupNS(t, r, "cs.cmu.edu")
+	fromParent := liveLookupNS(t, r, "cmu.edu")
+
+	assert.Equal(t, fromZone, fromHost)
+	assert.NotEqual(t, fromParent, fromHost)
+}
+
+// TestFindAuthoritativeNameservers_NoAnswer starts each walk for
+// www.google.com at 192.0.2.1, a documentation address where nothing
+// answers. A walk that got no answer does not say that the name has no
+// delegation of its own, so the lookup returns that walk's error, about
+// www.google.com, and tries no parent name: trying google.com and com
+// would end in ErrNoNameservers, or in the error of a walk for one of
+// them.
+func TestFindAuthoritativeNameservers_NoAnswer(t *testing.T) {
+	t.Parallel()
+
+	r := resolver.NewWithQueryTimeout(slog.Default(), 100*time.Millisecond)
+
+	nameservers, err := r.FindAuthoritativeNameserversFrom(
+		t.Context(), "www.google.com", []string{noAnswerAddress},
+	)
+	require.Error(t, err)
+	require.NotErrorIs(t, err, resolver.ErrNoNameservers)
+	assert.Contains(t, err.Error(), "query www.google.com. @"+noAnswerAddress)
+	assert.Empty(t, nameservers)
+}
+
 func TestFindAuthoritativeNameservers_ReturnsSorted(
 	t *testing.T,
 ) {
@@ -139,6 +190,134 @@ func TestFindAuthoritativeNameservers_CloudflareDomain(
 	}
 }
 
+// TestResolveNSIPs_EveryNameserver looks up the addresses of two of
+// google.com's nameservers together, as the walk does when a referral
+// names a zone's nameservers without their addresses, and compares them
+// with each looked up alone. Together they must give the addresses of
+// both, not only of the first that resolves, so that when one gives no
+// usable reply the walk goes on to the other.
+func TestResolveNSIPs_EveryNameserver(t *testing.T) {
+	t.Parallel()
+
+	r := newTestResolver(t)
+	names := []string{"ns3.google.com.", "ns4.google.com."}
+	want := make([]string, 0, len(names))
+
+	for _, name := range names {
+		want = append(want, liveResolveNSIPs(t, r, []string{name}, 1)...)
+	}
+
+	got := liveResolveNSIPs(t, r, names, len(want))
+
+	assert.ElementsMatch(t, want, got)
+}
+
+// TestResolveNSIPs_ZoneDelegatedWithoutAddresses looks up the address
+// of a.ntpns.org, a nameserver of pool.ntp.org. The org servers delegate
+// ntpns.org to nameservers in other zones and give none of their
+// addresses, so those are looked up on the way.
+func TestResolveNSIPs_ZoneDelegatedWithoutAddresses(t *testing.T) {
+	t.Parallel()
+
+	r := newTestResolver(t)
+	ips := liveResolveNSIPs(t, r, []string{"a.ntpns.org."}, 1)
+
+	for _, ip := range ips {
+		assert.NotNil(t, net.ParseIP(ip), "should be valid IP: %s", ip)
+	}
+}
+
+// TestQueryZone_GivenAddressesFail asks the servers of ntp.org about
+// pool.ntp.org, as the walk to a name under ntp.org does after the org
+// servers' referral. That referral names four nameservers and gives an
+// address for ns1.everett.org alone; here the given address is
+// 192.0.2.1, where nothing answers, so the other three must be looked
+// up and asked.
+func TestQueryZone_GivenAddressesFail(t *testing.T) {
+	t.Parallel()
+
+	r := newTestResolver(t)
+
+	var resp *dns.Msg
+
+	livednstest.Retry(
+		t,
+		"QueryZone(192.0.2.1 and three ntp.org nameservers, pool.ntp.org)",
+		func(ctx context.Context) error {
+			var err error
+
+			resp, err = r.QueryZone(
+				ctx, []string{"192.0.2.1"},
+				[]string{"anyns.pch.net.", "dns1.udel.edu.", "dns2.udel.edu."},
+				"ntp.org.", "pool.ntp.org.", dns.TypeNS, 0,
+			)
+
+			return err
+		},
+	)
+
+	assert.NotEmpty(t, resolver.NSSetFrom(resp, "pool.ntp.org."))
+}
+
+// TestQueryZone_LookupDepth asks the servers of g.ntpns.org, a
+// nameserver of pool.ntp.org, for its address, as looking that address
+// up does when anyns.pch.net, one of the servers of ntpns.org, gives the
+// referral to g.ntpns.org without addresses. Their addresses are looked
+// up (here only a.ntpns.org's), and that needs a bitnames.com
+// nameserver's address, as the org servers delegate ntpns.org without
+// addresses. From depth 1, where looking up g.ntpns.org's address
+// starts, that makes three lookups and the address is found. From one
+// below maxLookupDepth, the bitnames.com lookup would be past the limit,
+// so nothing can be asked, and the error says the limit is why.
+func TestQueryZone_LookupDepth(t *testing.T) {
+	t.Parallel()
+
+	r := newTestResolver(t)
+	withoutAddresses := []string{"a.ntpns.org."}
+
+	var resp *dns.Msg
+
+	livednstest.Retry(
+		t,
+		"QueryZone(a.ntpns.org without its address, g.ntpns.org)",
+		func(ctx context.Context) error {
+			var err error
+
+			resp, err = r.QueryZone(
+				ctx, nil, withoutAddresses, "g.ntpns.org.", "g.ntpns.org.",
+				dns.TypeA, 1,
+			)
+
+			return err
+		},
+	)
+
+	assert.NotEmpty(t, resp.Answer)
+
+	var limitErr error
+
+	// Any other error is live DNS not answering, and is retried.
+	livednstest.Retry(
+		t,
+		"QueryZone(a.ntpns.org without its address, g.ntpns.org, "+
+			"one below the limit)",
+		func(ctx context.Context) error {
+			_, limitErr = r.QueryZone(
+				ctx, nil, withoutAddresses, "g.ntpns.org.", "g.ntpns.org.",
+				dns.TypeA, resolver.MaxLookupDepth-1,
+			)
+			if limitErr == nil ||
+				errors.Is(limitErr, resolver.ErrLookupDepthExceeded) {
+				return nil
+			}
+
+			return limitErr
+		},
+	)
+
+	require.ErrorIs(t, limitErr, resolver.ErrLookupDepthExceeded)
+}
+
 // ----------------------------------------------------------------
 // QueryNameserver tests
 // ----------------------------------------------------------------
@@ -148,7 +327,7 @@ func TestQueryNameserver_BasicA(t *testing.T) {
 
 	r := newTestResolver(t)
 	ns := findOneNSForDomain(t, r, "google.com")
-	resp := liveQueryNameserver(t, r, ns, "www.google.com")
+	resp := liveQueryNameserver(t, r, ns, "www.google.com", "A", "CNAME")
 
 	require.NotNil(t, resp)
 
@@ -162,12 +341,26 @@ func TestQueryNameserver_BasicA(t *testing.T) {
 	)
 }
 
+// TestQueryNameserver_ZoneDelegatedWithoutAddresses asks a.ntpns.org, a
+// nameserver of pool.ntp.org, about pool.ntp.org, as the watcher does.
+// The org servers delegate ntpns.org without the addresses of its
+// nameservers, so finding a.ntpns.org's address needs a lookup inside
+// the one QueryNameserver starts.
+func TestQueryNameserver_ZoneDelegatedWithoutAddresses(t *testing.T) {
+	t.Parallel()
+
+	r := newTestResolver(t)
+	resp := liveQueryNameserver(t, r, "a.ntpns.org.", "pool.ntp.org", "A")
+
+	assert.Equal(t, resolver.StatusOK, resp.Status)
+}
+
 func TestQueryNameserver_AAAA(t *testing.T) {
 	t.Parallel()
 
 	r := newTestResolver(t)
 	ns := findOneNSForDomain(t, r, "cloudflare.com")
-	resp := liveQueryNameserver(t, r, ns, "cloudflare.com")
+	resp := liveQueryNameserver(t, r, ns, "cloudflare.com", "AAAA")
 
 	aaaaRecords := resp.Records["AAAA"]
 	require.NotEmpty(t, aaaaRecords,
@@ -187,7 +380,7 @@ func TestQueryNameserver_MX(t *testing.T) {
 
 	r := newTestResolver(t)
 	ns := findOneNSForDomain(t, r, "google.com")
-	resp := liveQueryNameserver(t, r, ns, "google.com")
+	resp := liveQueryNameserver(t, r, ns, "google.com", "MX")
 
 	mxRecords := resp.Records["MX"]
 	require.NotEmpty(t, mxRecords,
@@ -200,7 +393,7 @@ func TestQueryNameserver_TXT(t *testing.T) {
 
 	r := newTestResolver(t)
 	ns := findOneNSForDomain(t, r, "google.com")
-	resp := liveQueryNameserver(t, r, ns, "google.com")
+	resp := liveQueryNameserver(t, r, ns, "google.com", "TXT")
 
 	txtRecords := resp.Records["TXT"]
 	require.NotEmpty(t, txtRecords,
@@ -222,6 +415,30 @@ func TestQueryNameserver_TXT(t *testing.T) {
 	)
 }
 
+// TestQueryNameserver_TruncatedReplyWhoseTCPRetryFails asks a google.com
+// nameserver about google.com with a resolver whose retries over TCP
+// fail. google.com's TXT records do not fit in a reply over UDP, so TXT
+// is reported as failed, holding none of the records that fit, and
+// logged with the reason, while the nameserver, which answered the other
+// types, is ok.
+func TestQueryNameserver_TruncatedReplyWhoseTCPRetryFails(t *testing.T) {
+	t.Parallel()
+
+	ns := findOneNSForDomain(t, newTestResolver(t), "google.com")
+
+	var logs bytes.Buffer
+
+	r := resolver.NewWithFailingTCP(slog.New(slog.NewTextHandler(&logs, nil)))
+	resp := liveQueryNameserver(t, r, ns, "google.com")
+
+	assert.Equal(t, resolver.StatusOK, resp.Status)
+	assert.Contains(t, resp.FailedTypes, "TXT")
+	assert.NotContains(t, resp.Records, "TXT")
+	assert.Contains(t, logs.String(),
+		"hostname=google.com. nameserver="+ns+" type=TXT error=",
+	)
+}
+
 func TestQueryNameserver_NXDomain(t *testing.T) {
 	t.Parallel()
 
@@ -273,6 +490,184 @@ func TestQueryNameserver_Refused(t *testing.T) {
 	assert.Equal(t, "server returned REFUSED", resp.Error)
 }
 
+// TestQueryServers_RecursiveResolverRefused passes a public recursive
+// resolver to QueryServers as the server of google.com. These resolvers
+// refuse a query that does not ask for recursion and answer one that
+// does. The resolver never asks for recursion, so the query must be
+// reported as refused, never answered. Each resolver is run by a
+// different operator, and they are asked in turn until one replies, so
+// one operator not answering does not fail the test.
+func TestQueryServers_RecursiveResolverRefused(t *testing.T) {
+	t.Parallel()
+
+	r := newTestResolver(t)
+	resolvers := []string{
+		"64.6.64.6", "185.222.222.222", "4.2.2.1", "9.9.9.9",
+	}
+
+	var err error
+
+	livednstest.Retry(
+		t,
+		"QueryServers(public recursive resolvers, google.com)",
+		func(ctx context.Context) error {
+			for _, ip := range resolvers {
+				_, err = r.QueryServers(
+					ctx, []string{ip}, "google.com.", "google.com.",
+					dns.TypeA,
+				)
+
+				// A refusal or an answer is a reply; anything else may
+				// be no reply at all, so the next resolver is asked.
+				if err == nil || errors.Is(err, resolver.ErrRefused) {
+					return nil
+				}
+			}
+
+			return fmt.Errorf("%w: %w", livednstest.ErrNoAnswer, err)
+		},
+	)
+
+	require.ErrorIs(t, err, resolver.ErrRefused)
+}
+
+// googleNameserverIPv4s returns the IPv4 addresses of google.com's
+// nameservers, the only addresses the resolver asks servers at.
+func googleNameserverIPv4s(t *testing.T, r *resolver.Resolver) []string {
+	t.Helper()
+
+	names := liveFindAuthoritative(t, r, "google.com")
+
+	return liveResolveNSIPs(t, r, names, len(names))
+}
+
+// TestQueryServers_EveryServerRefused asks all of google.com's
+// nameservers about cloudflare.com, a zone they do not serve, which
+// they all refuse. The error says every server refused; it is not
+// ErrIntercepted, which only the root servers refusing shows.
+func TestQueryServers_EveryServerRefused(t *testing.T) {
+	t.Parallel()
+
+	r := newTestResolver(t)
+	servers := googleNameserverIPv4s(t, r)
+
+	var err error
+
+	livednstest.Retry(
+		t,
+		"QueryServers(google.com servers, cloudflare.com)",
+		func(ctx context.Context) error {
+			_, err = r.QueryServers(
+				ctx, servers, "google.com.", "cloudflare.com.",
+				dns.TypeNS,
+			)
+
+			// When not every server refused, one may have given no
+			// reply at all, so the attempt is tried again.
+			if err != nil &&
+				!strings.HasPrefix(err.Error(), "every server of") {
+				return fmt.Errorf(
+					"%w: %w", livednstest.ErrNoAnswer, err,
+				)
+			}
+
+			return nil
+		},
+	)
+
+	require.ErrorIs(t, err, resolver.ErrRefused)
+	require.NotErrorIs(t, err, resolver.ErrIntercepted)
+	require.EqualError(
+		t, err,
+		"every server of google.com. refused a query for "+
+			"cloudflare.com.: dns query refused",
+	)
+}
+
+// TestQueryServers_EveryRootServerRefused passes google.com's
+// nameservers to QueryServers as the servers of the root zone. They
+// refuse a query about cloudflare.com, as root servers would if
+// something on the network answered in their place, so the error is
+// ErrIntercepted.
+func TestQueryServers_EveryRootServerRefused(t *testing.T) {
+	t.Parallel()
+
+	r := newTestResolver(t)
+	servers := googleNameserverIPv4s(t, r)
+
+	var err error
+
+	livednstest.Retry(
+		t,
+		"QueryServers(google.com servers as root servers, cloudflare.com)",
+		func(ctx context.Context) error {
+			_, err = r.QueryServers(
+				ctx, servers, ".", "cloudflare.com.", dns.TypeNS,
+			)
+
+			// When not every server refused, one may have given no
+			// reply at all, so the attempt is tried again. Both errors
+			// for every server refusing say "refused a query for".
+			if err != nil &&
+				!strings.Contains(err.Error(), "refused a query for") {
+				return fmt.Errorf(
+					"%w: %w", livednstest.ErrNoAnswer, err,
+				)
+			}
+
+			return nil
+		},
+	)
+
+	require.ErrorIs(t, err, resolver.ErrIntercepted)
+	require.EqualError(
+		t, err,
+		"every root server refused a query for cloudflare.com.: "+
+			"this network intercepts DNS queries",
+	)
+}
+
+// TestQueryServers_NotEveryRootServerRefused passes google.com's
+// nameservers and 192.0.2.1 to QueryServers as the servers of the root
+// zone. The google.com nameservers refuse a query about cloudflare.com,
+// but nothing answers at 192.0.2.1, a documentation address, so not
+// every server refused, wherever 192.0.2.1 falls in the random order:
+// the error is not ErrIntercepted and does not say every server refused.
+func TestQueryServers_NotEveryRootServerRefused(t *testing.T) {
+	t.Parallel()
+
+	r := newTestResolver(t)
+	servers := googleNameserverIPv4s(t, r)
+	servers = append(servers, "192.0.2.1")
+
+	var err error
+
+	livednstest.Retry(
+		t,
+		"QueryServers(google.com servers and 192.0.2.1, cloudflare.com)",
+		func(ctx context.Context) error {
+			_, err = r.QueryServers(
+				ctx, servers, ".", "cloudflare.com.", dns.TypeNS,
+			)
+
+			// An attempt that ran out of time may not have asked every
+			// server, so it is tried again.
+			if ctx.Err() != nil {
+				return fmt.Errorf(
+					"%w: %w", livednstest.ErrNoAnswer, err,
+				)
+			}
+
+			return nil
+		},
+	)
+
+	require.Error(t, err)
+	require.NotErrorIs(t, err, resolver.ErrIntercepted)
+	// Both errors for every server refusing say "refused a query for".
+	require.NotContains(t, err.Error(), "refused a query for")
+}
+
 func TestQueryNameserver_RecordsSorted(t *testing.T) {
 	t.Parallel()
 
@@ -477,6 +872,145 @@ func TestLookupNS_MatchesFindAuthoritative(t *testing.T) {
 	assert.Equal(t, fromFind, fromLookup)
 }
 
+// TestLookupNS_ParentZoneDelegatedWithoutAddresses looks up the
+// nameservers of g.ntpns.org. The org servers delegate its parent zone,
+// ntpns.org, without the addresses of its nameservers, so the walk has
+// to look them up to ask them. If it did not, the walk for g.ntpns.org
+// would fail.
+func TestLookupNS_ParentZoneDelegatedWithoutAddresses(t *testing.T) {
+	t.Parallel()
+
+	r := newTestResolver(t)
+	nameservers := liveLookupNS(t, r, "g.ntpns.org")
+
+	assert.Contains(t, nameservers, "a.ntpns.org.")
+}
+
+// TestLookupNS_DomainThatDoesNotExist looks up the nameservers of a .com
+// domain that does not exist. The .com servers answer NXDOMAIN, so the
+// error is ErrNXDomain, and the domain does not get their names.
+func TestLookupNS_DomainThatDoesNotExist(t *testing.T) {
+	t.Parallel()
+
+	r := newTestResolver(t)
+
+	var (
+		nameservers []string
+		err         error
+	)
+
+	livednstest.Retry(
+		t,
+		"LookupNS("+nonexistentDomain+")",
+		func(ctx context.Context) error {
+			nameservers, err = r.LookupNS(ctx, nonexistentDomain)
+			if errors.Is(err, resolver.ErrNXDomain) {
+				return nil
+			}
+
+			return err
+		},
+	)
+
+	require.ErrorIs(t, err, resolver.ErrNXDomain)
+	assert.Empty(t, nameservers)
+}
+
+// TestLookupNS_NoDelegationOfItsOwn looks up the nameservers of
+// www.google.com, a name in the google.com zone with no delegation of
+// its own, as a domain such as octocat.github.io is. The google.com
+// servers answer with no NS records for it: the set is empty, and it is
+// not ErrNXDomain.
+func TestLookupNS_NoDelegationOfItsOwn(t *testing.T) {
+	t.Parallel()
+
+	r := newTestResolver(t)
+
+	var nameservers []string
+
+	livednstest.Retry(
+		t,
+		"LookupNS(www.google.com)",
+		func(ctx context.Context) error {
+			var err error
+
+			nameservers, err = r.LookupNS(ctx, "www.google.com")
+
+			return err
+		},
+	)
+
+	assert.Empty(t, nameservers)
+}
+
+// TestFollowDelegation_NoAnswer starts the walk LookupNS uses, for
+// google.com, at 192.0.2.1, a documentation address where nothing
+// answers. A walk that got no answer is an error, not an empty set,
+// which the watcher would report as an NS Change with every nameserver
+// removed.
+func TestFollowDelegation_NoAnswer(t *testing.T) {
+	t.Parallel()
+
+	r := resolver.NewWithQueryTimeout(slog.Default(), 100*time.Millisecond)
+
+	nameservers, err := r.FollowDelegation(
+		t.Context(), "google.com.", []string{noAnswerAddress},
+	)
+	require.Error(t, err)
+	assert.Empty(t, nameservers)
+}
+
+// TestResolveNSIterative_NoDelegationOfItsOwn walks to the nameservers
+// of www.google.com as the fallback walk does. As in
+// TestLookupNS_NoDelegationOfItsOwn, the set is empty, with no error.
+func TestResolveNSIterative_NoDelegationOfItsOwn(t *testing.T) {
+	t.Parallel()
+
+	r := newTestResolver(t)
+
+	var nameservers []string
+
+	livednstest.Retry(
+		t,
+		"ResolveNSIterative(www.google.com)",
+		func(ctx context.Context) error {
+			var err error
+
+			nameservers, err = r.ResolveNSIterative(ctx, "www.google.com")
+
+			return err
+		},
+	)
+
+	assert.Empty(t, nameservers)
+}
+
+// TestResolveNSIterative_DomainThatDoesNotExist walks to the nameservers
+// of a .com domain that does not exist as the fallback walk does. As in
+// TestLookupNS_DomainThatDoesNotExist, the error is ErrNXDomain.
+func TestResolveNSIterative_DomainThatDoesNotExist(t *testing.T) {
+	t.Parallel()
+
+	r := newTestResolver(t)
+
+	var err error
+
+	livednstest.Retry(
+		t,
+		"ResolveNSIterative("+nonexistentDomain+")",
+		func(ctx context.Context) error {
+			_, err = r.ResolveNSIterative(ctx, nonexistentDomain)
+			if errors.Is(err, resolver.ErrNXDomain) {
+				return nil
+			}
+
+			return err
+		},
+	)
+
+	require.ErrorIs(t, err, resolver.ErrNXDomain)
+}
+
 // ----------------------------------------------------------------
 // ResolveIPAddresses tests
 // ----------------------------------------------------------------
@@ -540,6 +1074,34 @@ func TestResolveIPAddresses_CloudflareDomain(t *testing.T) {
 	assert.NotEmpty(t, ips)
 }
 
+// TestResolveIPAddresses_NameserverIPv4AndIPv6 looks up the addresses of
+// one of cloudflare.com's nameservers, as a domain check does for each
+// nameserver. That name has A and AAAA records, so both kinds of address
+// come back.
+func TestResolveIPAddresses_NameserverIPv4AndIPv6(t *testing.T) {
+	t.Parallel()
+
+	r := newTestResolver(t)
+	ns := findOneNSForDomain(t, r, "cloudflare.com")
+	ips := liveResolveIPs(t, r, ns)
+
+	var ipv4, ipv6 int
+
+	for _, ip := range ips {
+		parsed := net.ParseIP(ip)
+		require.NotNil(t, parsed, "should be valid IP: %s", ip)
+
+		if parsed.To4() != nil {
+			ipv4++
+		} else {
+			ipv6++
+		}
+	}
+
+	assert.Positive(t, ipv4, "no IPv4 address for %s: %v", ns, ips)
+	assert.Positive(t, ipv6, "no IPv6 address for %s: %v", ns, ips)
+}
+
 // ----------------------------------------------------------------
 // Context cancellation tests
 // ----------------------------------------------------------------
@@ -633,6 +1195,105 @@ func TestQueryNameserverIP_Timeout(t *testing.T) {
 	assert.NotEmpty(t, resp.Error)
 }
 
+// TestQueryNameserverIP_CancelledLogsNothing cancels the context while
+// a query to 192.0.2.1, where nothing answers, is waiting for a reply,
+// as shutdown does. The query was cut short, not failed, so nothing is
+// logged.
+func TestQueryNameserverIP_CancelledLogsNothing(t *testing.T) {
+	t.Parallel()
+
+	var logs bytes.Buffer
+
+	r := resolver.NewFromLogger(slog.New(slog.NewTextHandler(&logs, nil)))
+
+	ctx, cancel := context.WithCancel(context.Background())
+	t.Cleanup(cancel)
+	time.AfterFunc(100*time.Millisecond, cancel)
+
+	_, err := r.QueryNameserverIP(
+		ctx, "unreachable.test.", "192.0.2.1", "example.com",
+	)
+	require.NoError(t, err)
+
+	assert.Empty(t, logs.String())
+}
+
+// TestCollectIPs_NoNameserverAnswered takes the response of a
+// nameserver at 192.0.2.1, where nothing answers, as
+// TestQueryNameserverIP_Timeout does. Addresses collected from
+// nameservers that all failed to answer are an error, not none.
+func TestCollectIPs_NoNameserverAnswered(t *testing.T) {
+	t.Parallel()
+
+	r := newTestResolver(t)
+
+	// The deadline outlasts the first try, as in
+	// TestQueryNameserverIP_Timeout.
+	ctx, cancel := context.WithTimeout(
+		context.Background(), 3*time.Second,
+	)
+	t.Cleanup(cancel)
+
+	resp, err := r.QueryNameserverIP(
+		ctx, "unreachable.test.", "192.0.2.1",
+		"example.com",
+	)
+	require.NoError(t, err)
+
+	ips, _, err := resolver.CollectIPs(
+		map[string]*resolver.NameserverResponse{resp.Nameserver: resp},
+	)
+	require.ErrorIs(t, err, resolver.ErrNoNameserverAnswered)
+	assert.Empty(t, ips)
+}
+
+// TestCollectIPs_ReferralIsNoAnswer asks a root server about
+// example.com, which the root zone does not hold, so it only refers the
+// query to the com servers. That reply is no answer, as is a parent
+// zone's when every server of the name's own zone failed.
+func TestCollectIPs_ReferralIsNoAnswer(t *testing.T) {
+	t.Parallel()
+
+	r := newTestResolver(t)
+
+	var resp *resolver.NameserverResponse
+
+	livednstest.Retry(
+		t,
+		"QueryNameserverIP(a.root-servers.net, example.com)",
+		func(ctx context.Context) error {
+			var err error
+
+			resp, err = r.QueryNameserverIP(
+				ctx, "a.root-servers.net.", "198.41.0.4",
+				"example.com",
+			)
+			if err != nil {
+				return err
+			}
+
+			// A timeout or a network error is no reply at all.
+			if resp.Status == resolver.StatusTimeout ||
+				strings.HasPrefix(resp.Error, "network error") {
+				return fmt.Errorf(
+					"%w: %s", livednstest.ErrNoAnswer, resp.Error,
+				)
+			}
+
+			return nil
+		},
+	)
+
+	assert.Equal(t, resolver.StatusError, resp.Status)
+	assert.Equal(t, "server returned a referral", resp.Error)
+
+	ips, _, err := resolver.CollectIPs(
+		map[string]*resolver.NameserverResponse{resp.Nameserver: resp},
+	)
+	require.ErrorIs(t, err, resolver.ErrNoNameserverAnswered)
+	assert.Empty(t, ips)
+}
+
 func TestResolveIPAddresses_ContextCanceled(t *testing.T) {
 	t.Parallel()
 

internal/state/state.go

@@ -8,6 +8,7 @@ import (
 	"log/slog"
 	"os"
 	"path/filepath"
+	"slices"
 	"sync"
 	"time"
 
@@ -37,24 +38,39 @@ type Params struct {
 // DomainState holds the monitoring state for an apex domain.
 // NameserverAddresses holds the sorted addresses each nameserver's name
 // resolves to, by nameserver name. A state file written before it
-// existed loads with it nil.
+// existed loads with it nil. NXDomain is true when the domain's parent
+// zone's servers answered that it does not exist; it then has no
+// nameservers.
 type DomainState struct {
 	Nameservers         []string            `json:"nameservers"`
 	NameserverAddresses map[string][]string `json:"nameserverAddresses"`
+	NXDomain            bool                `json:"nxdomain,omitempty"`
 	LastChecked         time.Time           `json:"lastChecked"`
 }
 
 // NameserverRecordState holds one NS's response for a hostname.
+// FailedTypes lists the record types whose query to the nameserver
+// failed on this check: Records holds for them the records saved by the
+// previous check, which are kept. UnknownTypes lists those of them whose
+// records the previous check did not know either, as when the
+// nameserver was new or failing then: Records holds nothing for them.
 type NameserverRecordState struct {
 	Records      map[string][]string `json:"records"`
+	FailedTypes  []string            `json:"failedTypes,omitempty"`
+	UnknownTypes []string            `json:"unknownTypes,omitempty"`
 	Status       string              `json:"status"`
 	Error        string              `json:"error,omitempty"`
 	LastChecked  time.Time           `json:"lastChecked"`
 }
 
 // HostnameState holds per-nameserver monitoring state for a hostname.
+// CNAMEAddresses holds the sorted addresses at the end of the name's
+// CNAME chain, found when its nameservers answered with a CNAME and no
+// address; it is empty otherwise. It is nil when they are not known: a
+// state file written before it existed loads with it nil.
 type HostnameState struct {
 	RecordsByNameserver map[string]*NameserverRecordState `json:"recordsByNameserver"`
+	CNAMEAddresses      []string                          `json:"cnameAddresses"`
 	LastChecked         time.Time                         `json:"lastChecked"`
 }
 
@@ -116,6 +132,8 @@ type CertificateState struct {
 }
 
 // Snapshot is the complete monitoring state persisted to disk.
+// Hostnames also holds each apex domain's own records, under the
+// domain's name, which has an entry in Domains too.
 type Snapshot struct {
 	Version      int                          `json:"version"`
 	LastUpdated  time.Time                    `json:"lastUpdated"`
@@ -196,6 +214,19 @@ func (s *State) Load() error {
 		return fmt.Errorf("parsing state file: %w", err)
 	}
 
+	// A state file saved before each record value was stored once can
+	// hold a hostname's CNAME once for every record type asked for.
+	// Each value is kept once, so the first check does not see a
+	// record change.
+	for _, hs := range snapshot.Hostnames {
+		for _, ns := range hs.RecordsByNameserver {
+			for recordType, values := range ns.Records {
+				slices.Sort(values)
+				ns.Records[recordType] = slices.Compact(values)
+			}
+		}
+	}
+
 	s.snapshot = &snapshot
 	s.log.Info("loaded state from disk", "path", path)
 
@@ -270,6 +301,27 @@ func (s *State) GetDomainState(
 	return ds, ok
 }
 
+// DeleteDomainState removes a domain state entry.
+func (s *State) DeleteDomainState(domain string) {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	delete(s.snapshot.Domains, domain)
+}
+
+// GetAllDomainNames returns the names of all domain state entries.
+func (s *State) GetAllDomainNames() []string {
+	s.mu.RLock()
+	defer s.mu.RUnlock()
+
+	names := make([]string, 0, len(s.snapshot.Domains))
+	for name := range s.snapshot.Domains {
+		names = append(names, name)
+	}
+
+	return names
+}
+
 // SetHostnameState updates the state for a hostname.
 func (s *State) SetHostnameState(
 	hostname string,
@@ -293,6 +345,28 @@ func (s *State) GetHostnameState(
 	return hs, ok
 }
 
+// DeleteHostnameState removes a hostname state entry.
+func (s *State) DeleteHostnameState(hostname string) {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	delete(s.snapshot.Hostnames, hostname)
+}
+
+// GetAllHostnames returns the names of all hostname state entries,
+// which include each apex domain's own records.
+func (s *State) GetAllHostnames() []string {
+	s.mu.RLock()
+	defer s.mu.RUnlock()
+
+	names := make([]string, 0, len(s.snapshot.Hostnames))
+	for name := range s.snapshot.Hostnames {
+		names = append(names, name)
+	}
+
+	return names
+}
+
 // SetPortState updates the state for a port.
 func (s *State) SetPortState(key string, ps *PortState) {
 	s.mu.Lock()
@@ -355,6 +429,27 @@ func (s *State) GetCertificateState(
 	return cs, ok
 }
 
+// DeleteCertificateState removes a certificate state entry.
+func (s *State) DeleteCertificateState(key string) {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	delete(s.snapshot.Certificates, key)
+}
+
+// GetAllCertificateKeys returns all certificate state keys.
+func (s *State) GetAllCertificateKeys() []string {
+	s.mu.RLock()
+	defer s.mu.RUnlock()
+
+	keys := make([]string, 0, len(s.snapshot.Certificates))
+	for k := range s.snapshot.Certificates {
+		keys = append(keys, k)
+	}
+
+	return keys
+}
+
 // checkDataDirWritable creates the data directory if needed, then writes
 // and removes the temp file that Save uses. It runs at startup so that an
 // unwritable directory stops the process, instead of the process running

internal/state/state_test.go

@@ -188,6 +188,203 @@ func TestLoadStateFromBeforeNameserverAddresses(t *testing.T) {
 	}
 }
 
+// TestSaveLoadRoundTrip_CNAMEAddresses checks that no addresses at the
+// end of a hostname's CNAME chain load as an empty list, and addresses
+// that are not known load as nil: the watcher tells the two apart.
+func TestSaveLoadRoundTrip_CNAMEAddresses(t *testing.T) {
+	t.Parallel()
+
+	dir := t.TempDir()
+	s := state.NewForTestWithDataDir(dir)
+
+	want := map[string][]string{
+		"cname.example.com":     {testIP},
+		"none.example.com":      {},
+		"not-known.example.com": nil,
+	}
+
+	for name, addresses := range want {
+		s.SetHostnameState(name, &state.HostnameState{
+			CNAMEAddresses: addresses,
+		})
+	}
+
+	err := s.Save()
+	if err != nil {
+		t.Fatalf("Save() error: %v", err)
+	}
+
+	loaded := state.NewForTestWithDataDir(dir)
+
+	err = loaded.Load()
+	if err != nil {
+		t.Fatalf("Load() error: %v", err)
+	}
+
+	for name, addresses := range want {
+		hs, ok := loaded.GetHostnameState(name)
+		if !ok {
+			t.Fatalf("missing hostname %s", name)
+		}
+
+		if !reflect.DeepEqual(hs.CNAMEAddresses, addresses) {
+			t.Errorf(
+				"%s: loaded %#v, want %#v",
+				name, hs.CNAMEAddresses, addresses,
+			)
+		}
+	}
+}
+
+// TestSaveLoadRoundTrip_FailedTypes checks that a nameserver's
+// failedTypes and unknownTypes survive a save and load. Without
+// unknownTypes, a type whose records were not known would load as one
+// with no records.
+func TestSaveLoadRoundTrip_FailedTypes(t *testing.T) {
+	t.Parallel()
+
+	dir := t.TempDir()
+	s := state.NewForTestWithDataDir(dir)
+
+	failed := []string{"TXT", "CAA"}
+	unknown := []string{"CAA"}
+
+	s.SetHostnameState(testHostname, &state.HostnameState{
+		RecordsByNameserver: map[string]*state.NameserverRecordState{
+			testNS1: {
+				Records:      map[string][]string{"TXT": {"v=spf1 -all"}},
+				FailedTypes:  failed,
+				UnknownTypes: unknown,
+				Status:       "ok",
+			},
+		},
+	})
+
+	err := s.Save()
+	if err != nil {
+		t.Fatalf("Save() error: %v", err)
+	}
+
+	loaded := state.NewForTestWithDataDir(dir)
+
+	err = loaded.Load()
+	if err != nil {
+		t.Fatalf("Load() error: %v", err)
+	}
+
+	hs, ok := loaded.GetHostnameState(testHostname)
+	if !ok {
+		t.Fatal("missing hostname " + testHostname)
+	}
+
+	ns1 := hs.RecordsByNameserver[testNS1]
+	if ns1 == nil {
+		t.Fatal("missing nameserver " + testNS1)
+	}
+
+	if !reflect.DeepEqual(ns1.FailedTypes, failed) {
+		t.Errorf("failedTypes: got %#v", ns1.FailedTypes)
+	}
+
+	if !reflect.DeepEqual(ns1.UnknownTypes, unknown) {
+		t.Errorf("unknownTypes: got %#v", ns1.UnknownTypes)
+	}
+}
+
+// TestLoadStateFromBeforeCNAMEAddresses loads a state file written
+// before the addresses at the end of a hostname's CNAME chain were
+// saved. They load as not known (nil), not as none.
+func TestLoadStateFromBeforeCNAMEAddresses(t *testing.T) {
+	t.Parallel()
+
+	dir := t.TempDir()
+
+	data := []byte(`{
+		"version": 1,
+		"lastUpdated": "2026-02-19T12:00:00Z",
+		"hostnames": {
+			"www.example.com": {
+				"recordsByNameserver": {},
+				"lastChecked": "2026-02-19T12:00:00Z"
+			}
+		}
+	}`)
+
+	err := os.WriteFile(filepath.Join(dir, "state.json"), data, 0o600)
+	if err != nil {
+		t.Fatalf("writing state file: %v", err)
+	}
+
+	s := state.NewForTestWithDataDir(dir)
+
+	err = s.Load()
+	if err != nil {
+		t.Fatalf("Load() error: %v", err)
+	}
+
+	hs, ok := s.GetHostnameState(testHostname)
+	if !ok {
+		t.Fatal("missing hostname " + testHostname)
+	}
+
+	if hs.CNAMEAddresses != nil {
+		t.Errorf("CNAME addresses: got %#v, want nil", hs.CNAMEAddresses)
+	}
+}
+
+// TestLoadStateWithRepeatedValues loads a state file saved when a
+// hostname's CNAME was stored once for every record type asked for.
+// Each value must load once, and every different value must load.
+func TestLoadStateWithRepeatedValues(t *testing.T) {
+	t.Parallel()
+
+	dir := t.TempDir()
+
+	data := []byte(`{
+		"version": 1,
+		"hostnames": {
+			"www.example.com": {
+				"recordsByNameserver": {
+					"ns1.example.com.": {
+						"records": {
+							"A": ["192.0.2.2", "192.0.2.1", "192.0.2.2", "192.0.2.1"],
+							"CNAME": ["a.example.net.", "a.example.net.", "a.example.net."]
+						},
+						"status": "ok"
+					}
+				}
+			}
+		}
+	}`)
+
+	err := os.WriteFile(filepath.Join(dir, "state.json"), data, 0o600)
+	if err != nil {
+		t.Fatalf("writing state file: %v", err)
+	}
+
+	s := state.NewForTestWithDataDir(dir)
+
+	err = s.Load()
+	if err != nil {
+		t.Fatalf("Load() error: %v", err)
+	}
+
+	hs, ok := s.GetHostnameState(testHostname)
+	if !ok {
+		t.Fatal("missing hostname " + testHostname)
+	}
+
+	want := map[string][]string{
+		"A":     {"192.0.2.1", "192.0.2.2"},
+		"CNAME": {"a.example.net."},
+	}
+
+	got := hs.RecordsByNameserver[testNS1].Records
+	if !reflect.DeepEqual(got, want) {
+		t.Errorf("records: got %v, want %v", got, want)
+	}
+}
+
 // TestSaveLoadRoundTrip_Hostnames verifies hostname data survives a save/load cycle.
 func TestSaveLoadRoundTrip_Hostnames(t *testing.T) {
 	t.Parallel()

internal/watcher/cancelled_test.go

@@ -1,10 +1,13 @@
 package watcher_test
 
 import (
+	"bytes"
 	"context"
 	"log/slog"
 	"reflect"
+	"strings"
 	"testing"
+	"time"
 
 	"sneak.berlin/go/dnswatcher/internal/portcheck"
 	"sneak.berlin/go/dnswatcher/internal/resolver"
@@ -79,3 +82,72 @@ func TestCancelledCheckSavesNothing(t *testing.T) {
 		t.Errorf("sent %v, want no notifications", notifications)
 	}
 }
+
+// newLoggingWatcher returns a watcher for a domain and a hostname, with
+// the real resolver, that writes what it logs at warning level or above
+// into the returned buffer.
+func newLoggingWatcher(t *testing.T) (*watcher.Watcher, *bytes.Buffer) {
+	t.Helper()
+
+	cfg := defaultTestConfig(t)
+	cfg.Domains = []string{testSmallDomain}
+	cfg.Hostnames = []string{host}
+
+	w, _ := newTestWatcher(t, cfg)
+
+	logs := &bytes.Buffer{}
+	w.SetLogger(slog.New(slog.NewJSONHandler(
+		logs, &slog.HandlerOptions{Level: slog.LevelWarn},
+	)))
+
+	return w, logs
+}
+
+// TestLookupCutShortIsNotLogged checks a domain and a hostname, looks
+// up a nameserver's addresses and follows a CNAME, with the context
+// cancelled, as shutdown leaves it. The real resolver fails each lookup
+// without sending a query. Shutdown cutting a lookup short is not a
+// failure, so nothing may be logged at warning level or above.
+func TestLookupCutShortIsNotLogged(t *testing.T) {
+	t.Parallel()
+
+	w, logs := newLoggingWatcher(t)
+
+	ctx, cancel := context.WithCancel(t.Context())
+	cancel()
+
+	w.RunOnce(ctx)
+	w.ResolveNameserverAddresses(ctx, []string{nsA}, nil)
+	w.ResolveCNAMEAddresses(ctx, host, cnameState(), nil)
+
+	if logs.Len() > 0 {
+		t.Errorf("logged at warning level or above:\n%s", logs)
+	}
+}
+
+// TestLookupOutOfTimeIsLoggedAsError does what
+// TestLookupCutShortIsNotLogged does, with the context's deadline passed
+// instead. A lookup that ran out of time did fail, so the domain's NS
+// lookup, the hostname's lookup, the nameserver's address lookup and the
+// CNAME's are each logged as an error.
+func TestLookupOutOfTimeIsLoggedAsError(t *testing.T) {
+	t.Parallel()
+
+	w, logs := newLoggingWatcher(t)
+
+	ctx, cancel := context.WithDeadline(t.Context(), time.Now())
+	t.Cleanup(cancel)
+
+	w.RunOnce(ctx)
+	w.ResolveNameserverAddresses(ctx, []string{nsA}, nil)
+	w.ResolveCNAMEAddresses(ctx, host, cnameState(), nil)
+
+	const want = 4
+
+	lines := strings.Count(logs.String(), "\n")
+	errorLines := strings.Count(logs.String(), `"level":"ERROR"`)
+
+	if lines != want || errorLines != want {
+		t.Errorf("logged:\n%s\nwant %d lines, each at error level", logs, want)
+	}
+}

internal/watcher/cname_test.go

@@ -0,0 +1,372 @@
+package watcher_test
+
+import (
+	"context"
+	"log/slog"
+	"slices"
+	"testing"
+
+	"sneak.berlin/go/dnswatcher/internal/livednstest"
+	"sneak.berlin/go/dnswatcher/internal/resolver"
+	"sneak.berlin/go/dnswatcher/internal/state"
+	"sneak.berlin/go/dnswatcher/internal/watcher"
+)
+
+// TestCNAMEIntoAnotherZonePortAndTLSChecks runs the port and TLS
+// checks on hostname state built here: the name's nameserver answered
+// with a CNAME into another zone, and following it found ip1. Both
+// checks must use ip1. They look nothing up, so the watcher has no
+// resolver.
+func TestCNAMEIntoAnotherZonePortAndTLSChecks(t *testing.T) {
+	t.Parallel()
+
+	cfg := defaultTestConfig(t)
+	cfg.Hostnames = []string{host}
+
+	deps := newTestDeps(t, cfg)
+	w := watcher.NewForTest(
+		cfg, deps.state, nil,
+		deps.portChecker, deps.tlsChecker, deps.notifier,
+	)
+
+	deps.state.SetHostnameState(host, cnameState(ip1))
+
+	w.CheckAllPorts(t.Context())
+	w.RunTLSChecks(t.Context())
+
+	snap := deps.state.GetSnapshot()
+
+	ps, ok := snap.Ports[ip1+":443"]
+	if !ok || !slices.Contains(ps.Hostnames, host) {
+		t.Errorf("no port state for %s at %s:443", host, ip1)
+	}
+
+	certKey := ip1 + ":443:" + host
+	if _, ok := snap.Certificates[certKey]; !ok {
+		t.Errorf("no certificate state %s", certKey)
+	}
+}
+
+// TestCNAMEThatCannotBeFollowedKeepsPrevious runs a check of a name, not
+// the watcher's first, from the point where its records have been looked
+// up: they hold a CNAME to a target under .invalid, whose lookup fails.
+// The previous check found the same records, and oldIP at the end of the
+// CNAME. The check must keep oldIP and send nothing.
+func TestCNAMEThatCannotBeFollowedKeepsPrevious(t *testing.T) {
+	t.Parallel()
+
+	w, deps := newTestWatcher(t, defaultTestConfig(t))
+	w.SetFirstRun(false)
+
+	records := map[string]map[string][]string{
+		nsA: cnameTo("target.example.invalid."),
+	}
+
+	prev := hostnameState(records)
+	prev.CNAMEAddresses = []string{oldIP}
+	deps.state.SetHostnameState(host, prev)
+
+	// The result is the same whether or not live DNS answers, so the
+	// lookup is not retried.
+	_ = livednstest.Run(func(ctx context.Context) error {
+		w.UpdateHostnameState(ctx, host, hostnameState(records))
+
+		return nil
+	})
+
+	hs, _ := deps.state.GetHostnameState(host)
+	if !slices.Equal(hs.CNAMEAddresses, prev.CNAMEAddresses) {
+		t.Errorf(
+			"saved %v, want %v",
+			hs.CNAMEAddresses, prev.CNAMEAddresses,
+		)
+	}
+
+	notifications := deps.notifier.getNotifications()
+	if len(notifications) != 0 {
+		t.Errorf("sent %v, want no notifications", notifications)
+	}
+}
+
+// followLive follows in live DNS the CNAMEs in a name's records, built
+// from records, and returns the addresses saved for the name. The
+// previous check saved oldIP, which is kept when a target cannot be
+// followed; that is retried. The tests point CNAMEs only at names in
+// zones with two nameservers, to keep queries few (see the top of
+// watcher_test.go).
+func followLive(
+	t *testing.T,
+	records map[string]map[string][]string,
+) []string {
+	t.Helper()
+
+	w := watcher.NewForTest(
+		nil, nil, resolver.NewFromLogger(slog.Default()), nil, nil, nil,
+	)
+	prev := cnameState(oldIP)
+
+	var current *state.HostnameState
+
+	livednstest.Retry(t, "following CNAMEs", func(ctx context.Context) error {
+		current = hostnameState(records)
+
+		w.ResolveCNAMEAddresses(ctx, host, current, prev)
+
+		if slices.Equal(current.CNAMEAddresses, prev.CNAMEAddresses) {
+			return livednstest.ErrNoAnswer
+		}
+
+		return nil
+	})
+
+	return current.CNAMEAddresses
+}
+
+// TestCNAMEAddressesOfEveryTarget gives a name's two nameservers
+// different CNAME targets, as when a secondary still serves an old one.
+// The addresses at the end of both are saved, whichever answer is read
+// first: one.one.one.one has 1.1.1.1, and dns.adguard-dns.com has
+// 94.140.14.14.
+func TestCNAMEAddressesOfEveryTarget(t *testing.T) {
+	t.Parallel()
+
+	found := followLive(t, map[string]map[string][]string{
+		nsA: cnameTo("one.one.one.one."),
+		nsB: cnameTo("dns.adguard-dns.com."),
+	})
+
+	for _, ip := range []string{"1.1.1.1", "94.140.14.14"} {
+		if !slices.Contains(found, ip) {
+			t.Errorf("saved %v, want %s among them", found, ip)
+		}
+	}
+}
+
+// TestCNAMEChainEndingInNoAddressSavesEmptyList follows a CNAME to a
+// name live DNS answers with NXDOMAIN. An empty list is saved, not nil,
+// which would mean the addresses are not known.
+func TestCNAMEChainEndingInNoAddressSavesEmptyList(t *testing.T) {
+	t.Parallel()
+
+	found := followLive(t, map[string]map[string][]string{
+		nsA: cnameTo("this-surely-does-not-exist-xyz.example.org."),
+	})
+
+	if found == nil || len(found) != 0 {
+		t.Errorf("saved %#v, want an empty list", found)
+	}
+}
+
+// TestCNAMEBesideAnAddressNotFollowed gives one nameserver of a name an
+// address and another a CNAME. The CNAME is not followed: an empty list
+// is saved, not nil, and nothing is looked up, the watcher having no
+// resolver.
+func TestCNAMEBesideAnAddressNotFollowed(t *testing.T) {
+	t.Parallel()
+
+	w := watcher.NewForTest(nil, nil, nil, nil, nil, nil)
+
+	current := hostnameState(map[string]map[string][]string{
+		nsA: {"A": {ip1}},
+		nsB: cnameTo("target.example.org."),
+	})
+
+	w.ResolveCNAMEAddresses(t.Context(), host, current, nil)
+
+	if current.CNAMEAddresses == nil || len(current.CNAMEAddresses) != 0 {
+		t.Errorf("saved %#v, want an empty list", current.CNAMEAddresses)
+	}
+}
+
+// TestCNAMEWhoseNameserversAllFailedKeepsPrevious checks a name none of
+// whose nameservers answered. The addresses the previous check saved
+// from following its CNAME are kept, and nothing is looked up: the
+// watcher has no resolver.
+func TestCNAMEWhoseNameserversAllFailedKeepsPrevious(t *testing.T) {
+	t.Parallel()
+
+	w := watcher.NewForTest(nil, nil, nil, nil, nil, nil)
+
+	current := saved(map[string]*state.NameserverRecordState{
+		nsA: failed(), nsB: failed(),
+	})
+	prev := cnameState(oldIP)
+
+	w.ResolveCNAMEAddresses(t.Context(), host, current, prev)
+
+	if !slices.Equal(current.CNAMEAddresses, prev.CNAMEAddresses) {
+		t.Errorf(
+			"saved %v, want %v",
+			current.CNAMEAddresses, prev.CNAMEAddresses,
+		)
+	}
+}
+
+// TestCNAMEWhoseAddressQueryFailedKeepsPrevious checks a name whose
+// nameserver answered, but whose query for A, AAAA or CNAME failed with
+// nothing kept for it. That is not an answer with no address: the
+// addresses the previous check saved from following its CNAME are kept,
+// and nothing is looked up, the watcher having no resolver.
+func TestCNAMEWhoseAddressQueryFailedKeepsPrevious(t *testing.T) {
+	t.Parallel()
+
+	for _, rtype := range []string{"A", "AAAA", "CNAME"} {
+		t.Run(rtype, func(t *testing.T) {
+			t.Parallel()
+
+			w := watcher.NewForTest(nil, nil, nil, nil, nil, nil)
+
+			current := saved(map[string]*state.NameserverRecordState{
+				nsA: {
+					Records:      map[string][]string{},
+					FailedTypes:  []string{rtype},
+					UnknownTypes: []string{rtype},
+					Status:       "ok",
+				},
+			})
+			prev := cnameState(oldIP)
+
+			w.ResolveCNAMEAddresses(t.Context(), host, current, prev)
+
+			if !slices.Equal(current.CNAMEAddresses, prev.CNAMEAddresses) {
+				t.Errorf(
+					"saved %v, want %v",
+					current.CNAMEAddresses, prev.CNAMEAddresses,
+				)
+			}
+		})
+	}
+}
+
+// cnameTo builds the records of a nameserver that answered with a CNAME
+// to target and no address.
+func cnameTo(target string) map[string][]string {
+	return map[string][]string{"CNAME": {target}}
+}
+
+// cnameState builds the state a check leaves behind for a name whose
+// nameserver answered with a CNAME and no address, when following the
+// CNAME found these addresses, which may be none.
+func cnameState(addresses ...string) *state.HostnameState {
+	hs := hostnameState(map[string]map[string][]string{
+		nsA: cnameTo("target.example.org."),
+	})
+
+	hs.CNAMEAddresses = append([]string{}, addresses...)
+
+	return hs
+}
+
+func TestCNAMEAddressChangeAlerts(t *testing.T) {
+	t.Parallel()
+
+	// A state file written before the addresses were saved loads with
+	// them nil.
+	olderStateFile := cnameState()
+	olderStateFile.CNAMEAddresses = nil
+
+	// Each case is the state saved by the previous check and by the
+	// current one. The name's records are the same in both.
+	tests := []struct {
+		name          string
+		prev, current *state.HostnameState
+		want          int
+	}{
+		{
+			"same addresses",
+			cnameState(ip1, ip2), cnameState(ip1, ip2), 0,
+		},
+		{
+			"same addresses in another order",
+			cnameState(ip2, ip1), cnameState(ip1, ip2), 0,
+		},
+		{
+			"address replaced",
+			cnameState(ip1), cnameState(ip2), 1,
+		},
+		{
+			"address added",
+			cnameState(ip1), cnameState(ip1, ip2), 1,
+		},
+		{
+			"no address at the end of the chain now",
+			cnameState(ip1), cnameState(), 1,
+		},
+		{
+			"addresses at the end of the chain again",
+			cnameState(), cnameState(ip1), 1,
+		},
+		{
+			"state file from before addresses were saved",
+			olderStateFile, cnameState(ip1), 0,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			notifier := &mockNotifier{}
+			w := watcher.NewForTest(nil, nil, nil, nil, nil, notifier)
+
+			w.DetectHostnameChanges(t.Context(), host, tt.prev, tt.current)
+
+			got := len(notifier.getNotifications())
+			if got != tt.want {
+				t.Errorf("sent %d notifications, want %d", got, tt.want)
+			}
+		})
+	}
+}
+
+func TestCNAMEAddressChangeAlertNamesHostnameAndAddresses(t *testing.T) {
+	t.Parallel()
+
+	notifier := &mockNotifier{}
+	w := watcher.NewForTest(nil, nil, nil, nil, nil, notifier)
+
+	w.DetectHostnameChanges(
+		t.Context(), host, cnameState(ip1), cnameState(ip2, ip3),
+	)
+
+	want := notification{
+		Title: "CNAME Address Change: " + host,
+		Message: "Hostname: " + host +
+			"\nOld: " + ip1 + "\nNew: " + ip2 + ", " + ip3,
+		Priority: "warning",
+	}
+
+	got := notifier.getNotifications()
+	if len(got) != 1 || got[0] != want {
+		t.Errorf("sent %v, want %v", got, want)
+	}
+}
+
+// TestNameMovedFromARecordsToCNAMEAlerts checks a name that answers
+// with an A record and then with a CNAME whose chain ends in ip2. The
+// second check is notified as a CNAME address change from no addresses,
+// beside the record change. Nothing is looked up: the watcher has no
+// resolver.
+func TestNameMovedFromARecordsToCNAMEAlerts(t *testing.T) {
+	t.Parallel()
+
+	notifier := &mockNotifier{}
+	w := watcher.NewForTest(nil, nil, nil, nil, nil, notifier)
+
+	prev := hostnameState(map[string]map[string][]string{
+		nsA: {"A": {ip1}},
+	})
+	w.ResolveCNAMEAddresses(t.Context(), host, prev, nil)
+
+	w.DetectHostnameChanges(t.Context(), host, prev, cnameState(ip2))
+
+	title := "CNAME Address Change: " + host
+	message := "Hostname: " + host + "\nOld: \nNew: " + ip2
+
+	got := notifier.getNotifications()
+	if !slices.ContainsFunc(got, func(n notification) bool {
+		return n.Title == title && n.Message == message
+	}) {
+		t.Errorf("sent %v, want %q with %q among them", got, title, message)
+	}
+}

internal/watcher/export_test.go

@@ -10,7 +10,8 @@ import (
 	"sneak.berlin/go/dnswatcher/internal/state"
 )
 
-// NewForTest creates a Watcher without fx for unit testing.
+// NewForTest creates a Watcher without fx for unit testing. A nil cfg
+// is an empty configuration.
 func NewForTest(
 	cfg *config.Config,
 	st *state.State,
@@ -19,6 +20,10 @@ func NewForTest(
 	tc TLSChecker,
 	n Notifier,
 ) *Watcher {
+	if cfg == nil {
+		cfg = &config.Config{}
+	}
+
 	return &Watcher{
 		log:       slog.Default(),
 		config:    cfg,
@@ -28,10 +33,15 @@ func NewForTest(
 		tlsCheck:  tc,
 		notify:    n,
 		firstRun:  true,
-		expiryNotified: make(map[string]time.Time),
 	}
 }
 
+// SetLogger replaces the watcher's logger, so a test can read what it
+// logs.
+func (w *Watcher) SetLogger(log *slog.Logger) {
+	w.log = log
+}
+
 // NewlyDisagreeingPairs exports newlyDisagreeingPairs for testing.
 func NewlyDisagreeingPairs(
 	prev, current *state.HostnameState,
@@ -39,6 +49,21 @@ func NewlyDisagreeingPairs(
 	return newlyDisagreeingPairs(prev, current)
 }
 
+// SetFirstRun sets whether the watcher is on its first check, in which
+// nothing is compared with the previous check. NewForTest's watcher is.
+func (w *Watcher) SetFirstRun(firstRun bool) {
+	w.firstRun = firstRun
+}
+
+// UpdateHostnameState exports updateHostnameState for testing.
+func (w *Watcher) UpdateHostnameState(
+	ctx context.Context,
+	hostname string,
+	newState *state.HostnameState,
+) {
+	w.updateHostnameState(ctx, hostname, newState)
+}
+
 // DetectHostnameChanges exports detectHostnameChanges for testing.
 func (w *Watcher) DetectHostnameChanges(
 	ctx context.Context,
@@ -58,6 +83,15 @@ func (w *Watcher) ResolveNameserverAddresses(
 	return w.resolveNameserverAddresses(ctx, nameservers, prev)
 }
 
+// ResolveCNAMEAddresses exports resolveCNAMEAddresses for testing.
+func (w *Watcher) ResolveCNAMEAddresses(
+	ctx context.Context,
+	hostname string,
+	current, prev *state.HostnameState,
+) {
+	w.resolveCNAMEAddresses(ctx, hostname, current, prev)
+}
+
 // DetectNSAddressChanges exports detectNSAddressChanges for testing.
 func (w *Watcher) DetectNSAddressChanges(
 	ctx context.Context,
@@ -67,10 +101,32 @@ func (w *Watcher) DetectNSAddressChanges(
 	w.detectNSAddressChanges(ctx, domain, prev, current)
 }
 
+// MaybeSendTestNotification exports maybeSendTestNotification for
+// testing.
+func (w *Watcher) MaybeSendTestNotification(ctx context.Context) {
+	w.maybeSendTestNotification(ctx)
+}
+
+// CleanupRemovedTargets exports cleanupRemovedTargets for testing.
+func (w *Watcher) CleanupRemovedTargets() {
+	w.cleanupRemovedTargets()
+}
+
+// CheckAllPorts exports checkAllPorts for testing.
+func (w *Watcher) CheckAllPorts(ctx context.Context) {
+	w.checkAllPorts(ctx)
+}
+
+// RunTLSChecks exports runTLSChecks for testing.
+func (w *Watcher) RunTLSChecks(ctx context.Context) {
+	w.runTLSChecks(ctx)
+}
+
 // BuildHostnameState exports buildHostnameState for testing.
 func BuildHostnameState(
 	results map[string]*resolver.NameserverResponse,
+	prev *state.HostnameState,
 	now time.Time,
 ) *state.HostnameState {
-	return buildHostnameState(results, now)
+	return buildHostnameState(results, prev, now)
 }

internal/watcher/failedtype_test.go

@@ -0,0 +1,364 @@
+package watcher_test
+
+import (
+	"maps"
+	"slices"
+	"testing"
+	"time"
+
+	"sneak.berlin/go/dnswatcher/internal/resolver"
+	"sneak.berlin/go/dnswatcher/internal/state"
+	"sneak.berlin/go/dnswatcher/internal/watcher"
+)
+
+const (
+	// txt is the record type whose query fails in these tests.
+	txt  = "TXT"
+	spf1 = "v=spf1 -all"
+	spf2 = "v=spf1 include:example.net -all"
+)
+
+// response is a nameserver's response with these records, whose queries
+// for failedTypes failed.
+func response(
+	records map[string][]string,
+	failedTypes ...string,
+) *resolver.NameserverResponse {
+	return &resolver.NameserverResponse{
+		Records:     records,
+		FailedTypes: failedTypes,
+		Status:      resolver.StatusOK,
+	}
+}
+
+// savedChecks saves the state of each check in turn from the
+// nameservers' responses, each from the state the check before saved.
+func savedChecks(
+	checks ...map[string]*resolver.NameserverResponse,
+) []*state.HostnameState {
+	states := make([]*state.HostnameState, 0, len(checks))
+
+	var prev *state.HostnameState
+
+	for _, results := range checks {
+		prev = watcher.BuildHostnameState(results, prev, time.Now())
+		states = append(states, prev)
+	}
+
+	return states
+}
+
+// TestFailedTypeKeepsPreviousRecords saves a check in which nsA's query
+// for TXT failed, after previous checks of several kinds. TXT is always
+// saved in FailedTypes, and in UnknownTypes when there was nothing to
+// keep.
+func TestFailedTypeKeepsPreviousRecords(t *testing.T) {
+	t.Parallel()
+
+	aOnly := map[string][]string{"A": {ip1}}
+	withTXT := map[string][]string{"A": {ip1}, txt: {spf1}}
+	txtKept := &state.NameserverRecordState{
+		Records: withTXT, FailedTypes: []string{txt}, Status: "ok",
+	}
+	txtNotKnown := &state.NameserverRecordState{
+		Records:      aOnly,
+		FailedTypes:  []string{txt},
+		UnknownTypes: []string{txt},
+		Status:       "ok",
+	}
+
+	tests := []struct {
+		name        string
+		prev        *state.HostnameState
+		wantRecords map[string][]string
+		wantUnknown []string
+	}{
+		{
+			"previous TXT records are kept",
+			saved(map[string]*state.NameserverRecordState{nsA: answered(withTXT)}),
+			withTXT, nil,
+		},
+		{
+			"previous check had no TXT records",
+			saved(map[string]*state.NameserverRecordState{nsA: answered(aOnly)}),
+			aOnly, nil,
+		},
+		{
+			"TXT failed on the previous check, which kept its records",
+			saved(map[string]*state.NameserverRecordState{nsA: txtKept}),
+			withTXT, nil,
+		},
+		{"first check", nil, aOnly, []string{txt}},
+		{
+			"nameserver new on this check",
+			saved(map[string]*state.NameserverRecordState{nsB: answered(withTXT)}),
+			aOnly, []string{txt},
+		},
+		{
+			"nameserver failed on the previous check",
+			saved(map[string]*state.NameserverRecordState{nsA: failed()}),
+			aOnly, []string{txt},
+		},
+		{
+			"TXT failed on the previous check with nothing to keep",
+			saved(map[string]*state.NameserverRecordState{nsA: txtNotKnown}),
+			aOnly, []string{txt},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			hs := watcher.BuildHostnameState(
+				map[string]*resolver.NameserverResponse{
+					nsA: response(map[string][]string{"A": {ip1}}, txt),
+				},
+				tt.prev, time.Now(),
+			)
+
+			got := hs.RecordsByNameserver[nsA]
+			if got.Status != "ok" ||
+				!maps.EqualFunc(got.Records, tt.wantRecords, slices.Equal) ||
+				!slices.Equal(got.FailedTypes, []string{txt}) ||
+				!slices.Equal(got.UnknownTypes, tt.wantUnknown) {
+				t.Errorf(
+					"saved status %q, records %v, failed types %v, "+
+						"unknown types %v; want ok, %v, [%s], %v",
+					got.Status, got.Records, got.FailedTypes,
+					got.UnknownTypes, tt.wantRecords, txt, tt.wantUnknown,
+				)
+			}
+		})
+	}
+}
+
+// TestFailedTypeAlerts saves the checks of each case in turn from the
+// nameservers' responses, the first being the state loaded at startup,
+// and counts the alerts sent. nsB's TXT query fails on one check, and
+// nothing changes.
+func TestFailedTypeAlerts(t *testing.T) {
+	t.Parallel()
+
+	records := map[string][]string{"A": {ip1}, txt: {spf1}}
+	aOnly := map[string][]string{"A": {ip1}}
+
+	bothAnswer := map[string]*resolver.NameserverResponse{
+		nsA: response(records), nsB: response(records),
+	}
+	bTXTFails := map[string]*resolver.NameserverResponse{
+		nsA: response(records), nsB: response(aOnly, txt),
+	}
+	onlyA := map[string]*resolver.NameserverResponse{
+		nsA: response(records),
+	}
+	bFails := map[string]*resolver.NameserverResponse{
+		nsA: response(records),
+		nsB: {
+			Records: map[string][]string{},
+			Status:  resolver.StatusTimeout,
+			Error:   "all queries timed out",
+		},
+	}
+
+	tests := []struct {
+		name   string
+		checks []map[string]*resolver.NameserverResponse
+		want   alertCounts
+	}{
+		{
+			"type failing at one nameserver alerts nothing, nor its next answer",
+			[]map[string]*resolver.NameserverResponse{
+				bothAnswer, bTXTFails, bothAnswer,
+			},
+			alertCounts{},
+		},
+		{
+			"type failing on the first check alerts nothing on the next",
+			[]map[string]*resolver.NameserverResponse{bTXTFails, bothAnswer},
+			alertCounts{},
+		},
+		{
+			"type failing at a nameserver new on that check alerts nothing",
+			[]map[string]*resolver.NameserverResponse{
+				onlyA, bTXTFails, bothAnswer,
+			},
+			alertCounts{},
+		},
+		{
+			"type failing at a recovering nameserver alerts the recovery",
+			[]map[string]*resolver.NameserverResponse{
+				bFails, bTXTFails, bothAnswer,
+			},
+			alertCounts{recoveries: 1},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			states := savedChecks(tt.checks...)
+
+			got := countAlerts(t, states[0], states[1:])
+			if got != tt.want {
+				t.Errorf("sent %+v, want %+v", got, tt.want)
+			}
+		})
+	}
+}
+
+// TestFailedTypeComparedOnceItAnswers saves the checks of each case in
+// turn as TestFailedTypeAlerts does. nsB's TXT query fails on one check,
+// and the TXT record changes: the change is sent as a Record Change for
+// each nameserver on the check where it answers it, and an Inconsistency
+// only when nsB still answers the old record.
+func TestFailedTypeComparedOnceItAnswers(t *testing.T) {
+	t.Parallel()
+
+	records := map[string][]string{"A": {ip1}, txt: {spf1}}
+	changed := map[string][]string{"A": {ip1}, txt: {spf2}}
+	aOnly := map[string][]string{"A": {ip1}}
+
+	bothAnswer := map[string]*resolver.NameserverResponse{
+		nsA: response(records), nsB: response(records),
+	}
+	bTXTFails := map[string]*resolver.NameserverResponse{
+		nsA: response(records), nsB: response(aOnly, txt),
+	}
+	bothChange := map[string]*resolver.NameserverResponse{
+		nsA: response(changed), nsB: response(changed),
+	}
+	aChangesBTXTFails := map[string]*resolver.NameserverResponse{
+		nsA: response(changed), nsB: response(aOnly, txt),
+	}
+	bStillOld := map[string]*resolver.NameserverResponse{
+		nsA: response(changed), nsB: response(records),
+	}
+
+	tests := []struct {
+		name   string
+		checks []map[string]*resolver.NameserverResponse
+		want   alertCounts
+	}{
+		{
+			"change made while the type failed",
+			[]map[string]*resolver.NameserverResponse{
+				bothAnswer, bTXTFails, bothChange,
+			},
+			alertCounts{recordChanges: 2},
+		},
+		{
+			"change seen at one nameserver while the other's type failed",
+			[]map[string]*resolver.NameserverResponse{
+				bothAnswer, aChangesBTXTFails, bothChange,
+			},
+			alertCounts{recordChanges: 2},
+		},
+		{
+			"old record answered after the type failed",
+			[]map[string]*resolver.NameserverResponse{
+				bothAnswer, aChangesBTXTFails, bStillOld,
+			},
+			alertCounts{recordChanges: 1, inconsistencies: 1},
+		},
+		{
+			"change after the type failed on the first check and answered",
+			[]map[string]*resolver.NameserverResponse{
+				bTXTFails, bothAnswer, bothChange,
+			},
+			alertCounts{recordChanges: 2},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			states := savedChecks(tt.checks...)
+
+			got := countAlerts(t, states[0], states[1:])
+			if got != tt.want {
+				t.Errorf("sent %+v, want %+v", got, tt.want)
+			}
+		})
+	}
+}
+
+// TestFailedTypeLeftOutOfMessages checks that a Record Change and an
+// Inconsistency name only the record types they compared. nsB's TXT
+// records are not known on the first check, and on the second either
+// answered or still not known; nsB's A record changes, so both alerts
+// are sent and name the A record alone.
+func TestFailedTypeLeftOutOfMessages(t *testing.T) {
+	t.Parallel()
+
+	withTXT := map[string][]string{"A": {ip1}, txt: {spf1}}
+	txtNotKnown := func(address string) *state.NameserverRecordState {
+		return &state.NameserverRecordState{
+			Records:      map[string][]string{"A": {address}},
+			FailedTypes:  []string{txt},
+			UnknownTypes: []string{txt},
+			Status:       "ok",
+		}
+	}
+
+	before := saved(map[string]*state.NameserverRecordState{
+		nsA: answered(withTXT), nsB: txtNotKnown(ip1),
+	})
+
+	tests := []struct {
+		name  string
+		after *state.HostnameState
+	}{
+		{
+			"TXT answers",
+			saved(map[string]*state.NameserverRecordState{
+				nsA: answered(withTXT),
+				nsB: answered(map[string][]string{"A": {ip2}, txt: {spf1}}),
+			}),
+		},
+		{
+			"TXT still not known",
+			saved(map[string]*state.NameserverRecordState{
+				nsA: answered(withTXT), nsB: txtNotKnown(ip2),
+			}),
+		},
+	}
+
+	want := map[string]string{
+		"Record Change: " + host: "Hostname: " + host +
+			"\nNameserver: " + nsB + "\nType: A\nOld: " + ip1 + "\nNew: " + ip2,
+		"Inconsistency: " + host: "Hostname: " + host +
+			"\nType: A\n" + nsA + ": " + ip1 + "\n" + nsB + ": " + ip2,
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			// The hostname change detection uses only the notifier.
+			notifier := &mockNotifier{}
+			w := watcher.NewForTest(nil, nil, nil, nil, nil, notifier)
+
+			w.DetectHostnameChanges(t.Context(), host, before, tt.after)
+
+			notifications := notifier.getNotifications()
+			if len(notifications) != len(want) {
+				t.Fatalf(
+					"sent %d notifications, want %d: %v",
+					len(notifications), len(want), notifications,
+				)
+			}
+
+			for _, n := range notifications {
+				if n.Message != want[n.Title] {
+					t.Errorf(
+						"%s message:\n%s\nwant:\n%s",
+						n.Title, n.Message, want[n.Title],
+					)
+				}
+			}
+		})
+	}
+}

internal/watcher/inconsistency_test.go

@@ -183,3 +183,58 @@ func TestInconsistencyAlert(t *testing.T) {
 		})
 	}
 }
+
+// TestFirstCheckAfterRepeatedValuesLoaded saves a state file holding a
+// hostname's CNAME once for every record type asked for, as checks did
+// before each value was stored once, and two addresses each repeated,
+// and loads it. A check that then finds each value once at each
+// nameserver must notify nothing.
+func TestFirstCheckAfterRepeatedValuesLoaded(t *testing.T) {
+	t.Parallel()
+
+	const (
+		cnameType = "CNAME"
+		cname     = "c.example.net."
+	)
+
+	cfg := defaultTestConfig(t)
+	repeated := map[string][]string{
+		"A":       {ip2, ip1, ip2, ip1},
+		cnameType: {cname, cname, cname, cname, cname, cname, cname, cname},
+	}
+	once := map[string][]string{"A": {ip1, ip2}, cnameType: {cname}}
+
+	saved := newTestDeps(t, cfg).state
+	saved.SetHostnameState(host, hostnameState(map[string]map[string][]string{
+		nsA: repeated, nsB: repeated,
+	}))
+
+	err := saved.Save()
+	if err != nil {
+		t.Fatalf("saving the state file: %v", err)
+	}
+
+	deps := newTestDeps(t, cfg)
+
+	err = deps.state.Load()
+	if err != nil {
+		t.Fatalf("loading the state file: %v", err)
+	}
+
+	prev, ok := deps.state.GetHostnameState(host)
+	if !ok {
+		t.Fatal("the state file has no state for " + host)
+	}
+
+	current := hostnameState(map[string]map[string][]string{
+		nsA: once, nsB: once,
+	})
+
+	// The hostname change detection uses only the notifier.
+	w := watcher.NewForTest(nil, nil, nil, nil, nil, deps.notifier)
+	w.DetectHostnameChanges(t.Context(), host, prev, current)
+
+	if got := deps.notifier.getNotifications(); len(got) != 0 {
+		t.Errorf("sent %v, want no notification", got)
+	}
+}

internal/watcher/interfaces.go

@@ -11,7 +11,9 @@ import (
 
 // DNSResolver performs iterative DNS resolution.
 type DNSResolver interface {
-	// LookupNS discovers authoritative nameservers for a domain.
+	// LookupNS returns a domain's NS record set, as its parent zone's
+	// servers delegate it: empty when they answer that it has none, and
+	// resolver.ErrNXDomain when they answer that it does not exist.
 	LookupNS(
 		ctx context.Context,
 		domain string,

internal/watcher/message_test.go

@@ -0,0 +1,215 @@
+package watcher_test
+
+import (
+	"maps"
+	"strings"
+	"testing"
+
+	"sneak.berlin/go/dnswatcher/internal/config"
+	"sneak.berlin/go/dnswatcher/internal/state"
+	"sneak.berlin/go/dnswatcher/internal/watcher"
+)
+
+// When one nameserver's A record changes and its TXT record does not,
+// the record change and the inconsistency it starts name the A record
+// alone, with its values written as plain text.
+func TestChangeMessagesNameTheChangedType(t *testing.T) {
+	t.Parallel()
+
+	// A nameserver's records: this A address and the same TXT record.
+	records := func(address string) map[string][]string {
+		return map[string][]string{
+			"A":   {address},
+			"TXT": {"v=spf1 -all"},
+		}
+	}
+
+	before := hostnameState(map[string]map[string][]string{
+		nsA: records(ip1),
+		nsB: records(ip1),
+	})
+	after := hostnameState(map[string]map[string][]string{
+		nsA: records(ip1),
+		nsB: records(ip2),
+	})
+
+	// The hostname change detection uses only the notifier.
+	notifier := &mockNotifier{}
+	w := watcher.NewForTest(nil, nil, nil, nil, nil, notifier)
+
+	w.DetectHostnameChanges(t.Context(), host, before, after)
+
+	want := map[string]string{
+		"Record Change: " + host: `Hostname: www.example.net
+Nameserver: b.ns.example.net.
+Type: A
+Old: 192.0.2.1
+New: 192.0.2.2`,
+		"Inconsistency: " + host: `Hostname: www.example.net
+Type: A
+a.ns.example.net.: 192.0.2.1
+b.ns.example.net.: 192.0.2.2`,
+	}
+
+	notifications := notifier.getNotifications()
+	if len(notifications) != len(want) {
+		t.Fatalf(
+			"sent %d notifications, want %d: %v",
+			len(notifications), len(want), notifications,
+		)
+	}
+
+	for _, n := range notifications {
+		if n.Message != want[n.Title] {
+			t.Errorf(
+				"%s message:\n%s\nwant:\n%s",
+				n.Title, n.Message, want[n.Title],
+			)
+		}
+	}
+}
+
+// Every kind of notification about a configured apex domain's own
+// records names it as a domain.
+func TestDomainRecordNotificationsNameTheDomain(t *testing.T) {
+	t.Parallel()
+
+	notifier := &mockNotifier{}
+	w := watcher.NewForTest(
+		&config.Config{Domains: []string{domain}},
+		nil, nil, nil, nil, notifier,
+	)
+
+	// nsA's address changes, which also makes it differ from nsC; nsB
+	// fails; nsC answers again; nsD is gone.
+	nsD := "d.ns.example.net."
+	w.DetectHostnameChanges(t.Context(), domain,
+		saved(map[string]*state.NameserverRecordState{
+			nsA: answered(map[string][]string{"A": {ip1}}),
+			nsB: answered(map[string][]string{"A": {ip1}}),
+			nsC: failed(),
+			nsD: answered(map[string][]string{"A": {ip1}}),
+		}),
+		saved(map[string]*state.NameserverRecordState{
+			nsA: answered(map[string][]string{"A": {ip2}}),
+			nsB: failed(),
+			nsC: answered(map[string][]string{"A": {ip1}}),
+		}),
+	)
+
+	// The address at the end of its CNAME chain changes.
+	w.DetectHostnameChanges(
+		t.Context(), domain, cnameState(ip1), cnameState(ip2),
+	)
+
+	// NS Failure is sent for nsB failing and for nsD being gone.
+	want := map[string]int{
+		"Record Change":        1,
+		"Inconsistency":        1,
+		"NS Failure":           2,
+		"NS Recovery":          1,
+		"CNAME Address Change": 1,
+	}
+
+	sent := make(map[string]int)
+
+	for _, n := range notifier.getNotifications() {
+		kind, _, _ := strings.Cut(n.Title, ":")
+		sent[kind]++
+
+		if !strings.HasPrefix(n.Message, "Domain: "+domain+"\n") {
+			t.Errorf("%s message does not name the domain:\n%s",
+				n.Title, n.Message)
+		}
+	}
+
+	if !maps.Equal(sent, want) {
+		t.Errorf("sent %v, want %v", sent, want)
+	}
+}
+
+// The startup notification counts the configured domains and hostnames,
+// although the state's hostnames also hold the apex domain's own
+// records. Nothing is looked up: the watcher has no resolver.
+func TestStartupNotificationCountsConfiguredNames(t *testing.T) {
+	t.Parallel()
+
+	cfg := defaultTestConfig(t)
+	cfg.SendTestNotification = true
+	cfg.Domains = []string{domain}
+	cfg.Hostnames = []string{host}
+
+	deps := newTestDeps(t, cfg)
+	w := watcher.NewForTest(cfg, deps.state, nil, nil, nil, deps.notifier)
+
+	// The state a check of both names saves.
+	deps.state.SetDomainState(domain, &state.DomainState{
+		Nameservers: []string{nsA},
+	})
+
+	for _, name := range []string{domain, host} {
+		deps.state.SetHostnameState(name, saved(
+			map[string]*state.NameserverRecordState{
+				nsA: answered(map[string][]string{"A": {ip1}}),
+			},
+		))
+	}
+
+	w.MaybeSendTestNotification(t.Context())
+
+	notifications := deps.notifier.getNotifications()
+
+	counts := "\nMonitoring 1 domain(s) and 1 hostname(s).\n"
+	if len(notifications) != 1 ||
+		!strings.Contains(notifications[0].Message, counts) {
+		t.Errorf("sent %v, want one message with %q", notifications, counts)
+	}
+}
+
+// A Port Change notification lists the configured apex domain and the
+// hostname that resolve to the port's address on separate lines. The
+// port checks read the saved hostname state and look nothing up, so the
+// watcher has no resolver.
+func TestPortChangeListsDomainsApartFromHostnames(t *testing.T) {
+	t.Parallel()
+
+	cfg := defaultTestConfig(t)
+	cfg.Domains = []string{domain}
+	cfg.Hostnames = []string{host}
+
+	deps := newTestDeps(t, cfg)
+	w := watcher.NewForTest(
+		cfg, deps.state, nil,
+		deps.portChecker, deps.tlsChecker, deps.notifier,
+	)
+	w.SetFirstRun(false)
+
+	// Both names resolve to ip1, whose port 443 the previous check
+	// found open. It is closed now.
+	for _, name := range []string{domain, host} {
+		deps.state.SetHostnameState(name, saved(
+			map[string]*state.NameserverRecordState{
+				nsA: answered(map[string][]string{"A": {ip1}}),
+			},
+		))
+	}
+
+	key := ip1 + ":443"
+	deps.state.SetPortState(key, &state.PortState{
+		Open: true, Hostnames: []string{domain, host},
+	})
+	deps.portChecker.closed = true
+
+	w.CheckAllPorts(t.Context())
+
+	title := "Port Change: " + key
+	want := `Domains: example.net
+Hostnames: www.example.net
+Address: 192.0.2.1:443
+Port now closed`
+
+	got := deps.notifier.getNotifications()
+	if len(got) != 1 || got[0].Title != title || got[0].Message != want {
+		t.Errorf("sent %v, want one %q with message:\n%s", got, title, want)
+	}
+}

internal/watcher/nsfailure_test.go

@@ -201,7 +201,7 @@ func TestNameserverThatNeverAnswers(t *testing.T) {
 	}
 
 	hs := watcher.BuildHostnameState(
-		map[string]*resolver.NameserverResponse{nsA: resp}, time.Now(),
+		map[string]*resolver.NameserverResponse{nsA: resp}, nil, time.Now(),
 	)
 
 	got := hs.RecordsByNameserver[nsA]
@@ -256,7 +256,7 @@ func TestNameserverThatAnswersNXDOMAIN(t *testing.T) {
 	}
 
 	hs := watcher.BuildHostnameState(
-		map[string]*resolver.NameserverResponse{ns: resp}, time.Now(),
+		map[string]*resolver.NameserverResponse{ns: resp}, nil, time.Now(),
 	)
 
 	got := hs.RecordsByNameserver[ns]
@@ -320,7 +320,7 @@ func TestNameserverThatRefuses(t *testing.T) {
 	}
 
 	hs := watcher.BuildHostnameState(
-		map[string]*resolver.NameserverResponse{ns: resp}, time.Now(),
+		map[string]*resolver.NameserverResponse{ns: resp}, nil, time.Now(),
 	)
 
 	got := hs.RecordsByNameserver[ns]
@@ -331,3 +331,118 @@ func TestNameserverThatRefuses(t *testing.T) {
 		)
 	}
 }
+
+// TestPortStateWhenNoNameserverAnswered runs the port checks on
+// hostname state built here, which gives the name no address. The port
+// state saved for its old address is kept only when the name is a
+// configured hostname or domain and none of its nameservers answered.
+func TestPortStateWhenNoNameserverAnswered(t *testing.T) {
+	t.Parallel()
+
+	noneAnswered := saved(map[string]*state.NameserverRecordState{
+		nsA: failed(), nsB: failed(),
+	})
+	oneAnsweredNoAddress := saved(map[string]*state.NameserverRecordState{
+		nsA: answered(map[string][]string{}), nsB: failed(),
+	})
+
+	configured := []string{host}
+
+	tests := []struct {
+		name      string
+		hostname  *state.HostnameState
+		hostnames []string
+		domains   []string
+		wantKept  bool
+	}{
+		{"no nameserver answered", noneAnswered, configured, nil, true},
+		{
+			"no nameserver answered, configured as a domain",
+			noneAnswered, nil, configured, true,
+		},
+		{
+			"one answered with no address",
+			oneAnsweredNoAddress, configured, nil, false,
+		},
+		{"no nameserver answered, not configured", noneAnswered, nil, nil, false},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			cfg := defaultTestConfig(t)
+			cfg.Hostnames = tt.hostnames
+			cfg.Domains = tt.domains
+
+			// The port checks read the saved hostname state and look
+			// nothing up, so the watcher has no resolver.
+			deps := newTestDeps(t, cfg)
+			w := watcher.NewForTest(
+				cfg, deps.state, nil,
+				deps.portChecker, deps.tlsChecker, deps.notifier,
+			)
+
+			key := ip1 + ":443"
+
+			deps.state.SetHostnameState(host, tt.hostname)
+			deps.state.SetPortState(key, &state.PortState{
+				Open: true, Hostnames: []string{host},
+			})
+
+			w.CheckAllPorts(t.Context())
+
+			_, kept := deps.state.GetPortState(key)
+			if kept != tt.wantKept {
+				t.Errorf("port state %s kept: %v, want %v", key, kept, tt.wantKept)
+			}
+		})
+	}
+}
+
+// TestPortStateWhenNoNameserverAnsweredAndOtherNameMovesAway saves the
+// port state of an address two configured hostnames resolve to. While
+// none of the first one's nameservers answer, the port checks run with
+// the other one still at that address, then after it moved away; the
+// port state is kept both times.
+func TestPortStateWhenNoNameserverAnsweredAndOtherNameMovesAway(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	const other = "mail.example.net"
+
+	cfg := defaultTestConfig(t)
+	cfg.Hostnames = []string{host, other}
+
+	// The port checks read the saved hostname state and look nothing
+	// up, so the watcher has no resolver.
+	deps := newTestDeps(t, cfg)
+	w := watcher.NewForTest(
+		cfg, deps.state, nil,
+		deps.portChecker, deps.tlsChecker, deps.notifier,
+	)
+
+	key := ip1 + ":443"
+
+	deps.state.SetPortState(key, &state.PortState{
+		Open: true, Hostnames: []string{host, other},
+	})
+	deps.state.SetHostnameState(host, saved(
+		map[string]*state.NameserverRecordState{nsA: failed(), nsB: failed()},
+	))
+
+	for _, otherIP := range []string{ip1, ip2} {
+		deps.state.SetHostnameState(other, saved(
+			map[string]*state.NameserverRecordState{
+				nsA: answered(map[string][]string{"A": {otherIP}}),
+			},
+		))
+
+		w.CheckAllPorts(t.Context())
+
+		if _, kept := deps.state.GetPortState(key); !kept {
+			t.Fatalf("port state %s removed with %s at %s", key, other, otherIP)
+		}
+	}
+}

internal/watcher/removed_test.go

@@ -0,0 +1,224 @@
+package watcher_test
+
+import (
+	"maps"
+	"slices"
+	"testing"
+
+	"sneak.berlin/go/dnswatcher/internal/state"
+	"sneak.berlin/go/dnswatcher/internal/watcher"
+)
+
+// TestRemovedTargetsLeaveTheState loads a state saved while a domain
+// and a hostname now removed from the configuration were still in it,
+// and runs the removal that Run does before the first check. The
+// removed names' domain, hostname and certificate entries are gone,
+// the configured names' are kept, and nothing is notified. Nothing is
+// looked up: the watcher has no resolver.
+func TestRemovedTargetsLeaveTheState(t *testing.T) {
+	t.Parallel()
+
+	const (
+		removedDomain = "example.com"
+		removedHost   = "www.example.com"
+	)
+
+	cfg := defaultTestConfig(t)
+	cfg.Domains = []string{domain}
+	cfg.Hostnames = []string{host}
+
+	deps := newTestDeps(t, cfg)
+	w := watcher.NewForTest(
+		cfg, deps.state, nil,
+		deps.portChecker, deps.tlsChecker, deps.notifier,
+	)
+
+	// The state a check of all four names saves, each name at ip1.
+	for _, name := range []string{domain, removedDomain} {
+		deps.state.SetDomainState(name, &state.DomainState{
+			Nameservers: []string{nsA},
+		})
+	}
+
+	for _, name := range []string{domain, host, removedDomain, removedHost} {
+		deps.state.SetHostnameState(name, saved(
+			map[string]*state.NameserverRecordState{
+				nsA: answered(map[string][]string{"A": {ip1}}),
+			},
+		))
+		deps.state.SetCertificateState(
+			ip1+":443:"+name, &state.CertificateState{Status: "ok"},
+		)
+	}
+
+	err := deps.state.Save()
+	if err != nil {
+		t.Fatalf("saving the state: %v", err)
+	}
+
+	err = deps.state.Load()
+	if err != nil {
+		t.Fatalf("loading the state: %v", err)
+	}
+
+	w.CleanupRemovedTargets()
+
+	snap := deps.state.GetSnapshot()
+
+	got := slices.Sorted(maps.Keys(snap.Domains))
+	if want := []string{domain}; !slices.Equal(got, want) {
+		t.Errorf("domain entries %v, want %v", got, want)
+	}
+
+	got = slices.Sorted(maps.Keys(snap.Hostnames))
+	if want := []string{domain, host}; !slices.Equal(got, want) {
+		t.Errorf("hostname entries %v, want %v", got, want)
+	}
+
+	got = slices.Sorted(maps.Keys(snap.Certificates))
+	if want := []string{
+		ip1 + ":443:" + domain, ip1 + ":443:" + host,
+	}; !slices.Equal(got, want) {
+		t.Errorf("certificate entries %v, want %v", got, want)
+	}
+
+	if sent := deps.notifier.getNotifications(); len(sent) != 0 {
+		t.Errorf("sent %v, want nothing", sent)
+	}
+}
+
+// TestRemovedTargetsLeaveThePortEntries loads a state whose port
+// entries name a domain and a hostname now removed from the
+// configuration, and runs the removal that Run does before the first
+// check. The removed names are off each port entry's list of names, the
+// entry only they had is gone, the entry that also names configured
+// names is kept for the port checks, and nothing is notified.
+func TestRemovedTargetsLeaveThePortEntries(t *testing.T) {
+	t.Parallel()
+
+	const (
+		removedDomain = "example.com"
+		removedHost   = "www.example.com"
+	)
+
+	cfg := defaultTestConfig(t)
+	cfg.Domains = []string{domain}
+	cfg.Hostnames = []string{host}
+
+	deps := newTestDeps(t, cfg)
+	w := watcher.NewForTest(
+		cfg, deps.state, nil,
+		deps.portChecker, deps.tlsChecker, deps.notifier,
+	)
+
+	// The port 443 entries a check of all four names saves: each name
+	// at ip1, except the removed hostname, at ip2.
+	deps.state.SetPortState(ip1+":443", &state.PortState{
+		Open: true, Hostnames: []string{removedDomain, domain, host},
+	})
+	deps.state.SetPortState(ip2+":443", &state.PortState{
+		Open: true, Hostnames: []string{removedHost},
+	})
+
+	err := deps.state.Save()
+	if err != nil {
+		t.Fatalf("saving the state: %v", err)
+	}
+
+	err = deps.state.Load()
+	if err != nil {
+		t.Fatalf("loading the state: %v", err)
+	}
+
+	w.CleanupRemovedTargets()
+
+	if sent := deps.notifier.getNotifications(); len(sent) != 0 {
+		t.Errorf("sent %v, want nothing", sent)
+	}
+
+	snap := deps.state.GetSnapshot()
+
+	got := slices.Sorted(maps.Keys(snap.Ports))
+	if want := []string{ip1 + ":443"}; !slices.Equal(got, want) {
+		t.Fatalf("port entries %v, want %v", got, want)
+	}
+
+	got = snap.Ports[ip1+":443"].Hostnames
+	if want := []string{domain, host}; !slices.Equal(got, want) {
+		t.Errorf("names of port entry %s:443 %v, want %v", ip1, got, want)
+	}
+}
+
+// TestCertificateStateForAnAddressGone runs the port checks on hostname
+// state built here for a configured hostname, with certificate entries
+// saved for it at ip1, ip2 and an IPv6 address. When its nameservers
+// answered with ip1 and the IPv6 address, the entry for ip2 is removed.
+// When none of them answered, its addresses are not known, and every
+// entry is kept. Nothing is notified, and nothing is looked up.
+func TestCertificateStateForAnAddressGone(t *testing.T) {
+	t.Parallel()
+
+	const ip6 = "2001:db8::1"
+
+	tests := []struct {
+		name     string
+		hostname *state.HostnameState
+		want     []string
+	}{
+		{
+			"answered without ip2",
+			saved(map[string]*state.NameserverRecordState{
+				nsA: answered(map[string][]string{
+					"A": {ip1}, "AAAA": {ip6},
+				}),
+			}),
+			[]string{ip1 + ":443:" + host, ip6 + ":443:" + host},
+		},
+		{
+			"no nameserver answered",
+			saved(map[string]*state.NameserverRecordState{
+				nsA: failed(), nsB: failed(),
+			}),
+			[]string{
+				ip1 + ":443:" + host,
+				ip2 + ":443:" + host,
+				ip6 + ":443:" + host,
+			},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			cfg := defaultTestConfig(t)
+			cfg.Hostnames = []string{host}
+
+			deps := newTestDeps(t, cfg)
+			w := watcher.NewForTest(
+				cfg, deps.state, nil,
+				deps.portChecker, deps.tlsChecker, deps.notifier,
+			)
+			w.SetFirstRun(false)
+
+			deps.state.SetHostnameState(host, tt.hostname)
+
+			for _, ip := range []string{ip1, ip2, ip6} {
+				deps.state.SetCertificateState(
+					ip+":443:"+host, &state.CertificateState{Status: "ok"},
+				)
+			}
+
+			w.CheckAllPorts(t.Context())
+
+			got := slices.Sorted(maps.Keys(deps.state.GetSnapshot().Certificates))
+			if !slices.Equal(got, tt.want) {
+				t.Errorf("certificate entries %v, want %v", got, tt.want)
+			}
+
+			if sent := deps.notifier.getNotifications(); len(sent) != 0 {
+				t.Errorf("sent %v, want nothing", sent)
+			}
+		})
+	}
+}

internal/watcher/watcher.go

@@ -2,11 +2,13 @@ package watcher
 
 import (
 	"context"
+	"errors"
 	"fmt"
 	"log/slog"
+	"maps"
+	"slices"
 	"sort"
 	"strings"
-	"sync"
 	"time"
 
 	"go.uber.org/fx"
@@ -58,8 +60,6 @@ type Watcher struct {
 	cancel    context.CancelFunc
 	done      chan struct{} // closed when Run returns
 	firstRun  bool
-	expiryNotifiedMu sync.Mutex
-	expiryNotified   map[string]time.Time
 }
 
 // New creates a new Watcher instance wired into the fx lifecycle.
@@ -76,7 +76,6 @@ func New(
 		tlsCheck:  params.TLSCheck,
 		notify:    params.Notify,
 		firstRun:  true,
-		expiryNotified: make(map[string]time.Time),
 	}
 
 	lifecycle.Append(fx.Hook{
@@ -126,10 +125,13 @@ func (w *Watcher) Run(ctx context.Context) {
 		"watcher starting",
 		"domains", len(w.config.Domains),
 		"hostnames", len(w.config.Hostnames),
-		"dnsInterval", w.config.DNSInterval,
+		// As text: the JSON log writes a time.Duration as bare
-		"tlsInterval", w.config.TLSInterval,
+		// nanoseconds.
+		"dnsInterval", w.config.DNSInterval.String(),
+		"tlsInterval", w.config.TLSInterval.String(),
 	)
 
+	w.cleanupRemovedTargets()
 	w.RunOnce(ctx)
 	w.maybeSendTestNotification(ctx)
 
@@ -200,6 +202,59 @@ func (w *Watcher) detectFirstRun() {
 	}
 }
 
+// cleanupRemovedTargets removes from the loaded state the domain,
+// hostname and certificate entries of names no longer in the
+// configuration, which changes only at a restart, and takes those names
+// off each port entry's list of names, removing a port entry left with
+// none. Nothing is notified. A configured domain's own records are
+// saved as a hostname entry under its name, which is kept.
+func (w *Watcher) cleanupRemovedTargets() {
+	for _, name := range w.state.GetAllDomainNames() {
+		if !w.isDomain(name) {
+			w.state.DeleteDomainState(name)
+		}
+	}
+
+	for _, name := range w.state.GetAllHostnames() {
+		if !w.isConfigured(name) {
+			w.state.DeleteHostnameState(name)
+		}
+	}
+
+	for _, key := range w.state.GetAllCertificateKeys() {
+		if _, hostname := parseCertKey(key); !w.isConfigured(hostname) {
+			w.state.DeleteCertificateState(key)
+		}
+	}
+
+	for _, key := range w.state.GetAllPortKeys() {
+		ps, ok := w.state.GetPortState(key)
+		if !ok {
+			continue
+		}
+
+		var names []string
+
+		for _, name := range ps.Hostnames {
+			if w.isConfigured(name) {
+				names = append(names, name)
+			}
+		}
+
+		if len(names) == 0 {
+			w.state.DeletePortState(key)
+
+			continue
+		}
+
+		w.state.SetPortState(key, &state.PortState{
+			Open:        ps.Open,
+			Hostnames:   names,
+			LastChecked: ps.LastChecked,
+		})
+	}
+}
+
 // runDNSChecks performs DNS resolution for all configured domains
 // and hostnames, updating state with freshly resolved records.
 // This must complete before port or TLS checks run so those
@@ -214,13 +269,36 @@ func (w *Watcher) runDNSChecks(ctx context.Context) {
 	}
 }
 
+// logFailedLookup logs a failed DNS lookup at error level, unless ctx
+// was cancelled: shutdown cancels it, and a lookup it cut short did not
+// fail. A lookup that ran out of time did fail, so it is logged.
+func (w *Watcher) logFailedLookup(
+	ctx context.Context,
+	msg string,
+	args ...any,
+) {
+	if errors.Is(ctx.Err(), context.Canceled) {
+		return
+	}
+
+	w.log.Error(msg, args...)
+}
+
 func (w *Watcher) checkDomain(
 	ctx context.Context,
 	domain string,
 ) {
 	nameservers, err := w.resolver.LookupNS(ctx, domain)
+
+	// A domain that does not exist has no nameservers.
+	nxdomain := errors.Is(err, resolver.ErrNXDomain)
+	if nxdomain {
+		nameservers, err = []string{}, nil
+	}
+
 	if err != nil {
-		w.log.Error(
+		w.logFailedLookup(
+			ctx,
 			"failed to lookup NS",
 			"domain", domain,
 			"error", err,
@@ -252,31 +330,22 @@ func (w *Watcher) checkDomain(
 	w.state.SetDomainState(domain, &state.DomainState{
 		Nameservers:         nameservers,
 		NameserverAddresses: addresses,
+		NXDomain:            nxdomain,
 		LastChecked:         now,
 	})
 
-	// Also look up A/AAAA records for the apex domain so that
+	// A domain that does not exist has no records of its own: none are
-	// port and TLS checks (which read HostnameState) can find
+	// asked for, and those saved by an earlier check are removed.
-	// the domain's IP addresses.
+	if nxdomain {
-	results, err := w.resolver.LookupAllRecords(ctx, domain)
+		w.state.DeleteHostnameState(domain)
-	if err != nil {
-		w.log.Error(
-			"failed to lookup records for domain",
-			"domain", domain,
-			"error", err,
-		)
 
 		return
 	}
 
-	newState := buildHostnameState(results, now)
+	// The apex domain's records are also checked and saved as a
-
+	// hostname's, so that the port and TLS checks find its addresses.
-	prevHS, hasPrevHS := w.state.GetHostnameState(domain)
+	// Notifications about them name it as a domain (see nameLine).
-	if hasPrevHS && !w.firstRun {
+	w.checkHostname(ctx, domain)
-		w.detectHostnameChanges(ctx, domain, prevHS, newState)
-	}
-
-	w.state.SetHostnameState(domain, newState)
 }
 
 func (w *Watcher) detectNSChanges(
@@ -321,10 +390,9 @@ func (w *Watcher) detectNSChanges(
 }
 
 // resolveNameserverAddresses returns the sorted addresses each
-// nameserver's name resolves to. A nameserver whose lookup fails or
+// nameserver's name resolves to. A nameserver whose lookup fails, as it
-// finds no address keeps its addresses from prev: the resolver finds no
+// does when no nameserver of the name's zone answers, or finds no
-// address, without an error, when every server it asks times out, and
+// address keeps its addresses from prev and is not an address change.
-// that is not an address change.
 func (w *Watcher) resolveNameserverAddresses(
 	ctx context.Context,
 	nameservers []string,
@@ -341,7 +409,8 @@ func (w *Watcher) resolveNameserverAddresses(
 			continue
 		}
 
-		w.log.Error(
+		w.logFailedLookup(
+			ctx,
 			"no addresses found for nameserver",
 			"nameserver", ns,
 			"error", err,
@@ -393,7 +462,8 @@ func (w *Watcher) checkHostname(
 ) {
 	results, err := w.resolver.LookupAllRecords(ctx, hostname)
 	if err != nil {
-		w.log.Error(
+		w.logFailedLookup(
+			ctx,
 			"failed to lookup records",
 			"hostname", hostname,
 			"error", err,
@@ -402,9 +472,25 @@ func (w *Watcher) checkHostname(
 		return
 	}
 
-	newState := buildHostnameState(results, time.Now().UTC())
+	prev, _ := w.state.GetHostnameState(hostname)
 
+	w.updateHostnameState(
+		ctx, hostname, buildHostnameState(results, prev, time.Now().UTC()),
+	)
+}
+
+// updateHostnameState finishes a check of hostname from newState, built
+// from its nameservers' answers: it follows the CNAME in them, notifies
+// what changed since the previous check, and saves newState.
+func (w *Watcher) updateHostnameState(
+	ctx context.Context,
+	hostname string,
+	newState *state.HostnameState,
+) {
 	prev, hasPrev := w.state.GetHostnameState(hostname)
+
+	w.resolveCNAMEAddresses(ctx, hostname, newState, prev)
+
 	if hasPrev && !w.firstRun {
 		w.detectHostnameChanges(ctx, hostname, prev, newState)
 	}
@@ -412,12 +498,89 @@ func (w *Watcher) checkHostname(
 	w.state.SetHostnameState(hostname, newState)
 }
 
+// resolveCNAMEAddresses saves in current the addresses at the end of
+// hostname's CNAME chain, when the nameservers' answers in current hold
+// a CNAME and no address, and an empty list otherwise. Every CNAME
+// target the nameservers gave is followed with ResolveIPAddresses and
+// the addresses found for all of them are saved, so nameservers that
+// disagree on the target do not change the result from check to check.
+// The addresses saved in prev, which may be nil, are kept when none of
+// the name's nameservers answered its queries for A, AAAA and CNAME,
+// and when a target cannot be followed, as when no nameserver of a zone
+// in its chain answers.
+func (w *Watcher) resolveCNAMEAddresses(
+	ctx context.Context,
+	hostname string,
+	current, prev *state.HostnameState,
+) {
+	var prevAddresses []string
+	if prev != nil {
+		prevAddresses = prev.CNAMEAddresses
+	}
+
+	// Empty, not nil: nil means the addresses are not known.
+	current.CNAMEAddresses = []string{}
+
+	answered := false
+	targets := make(map[string]bool)
+
+	for _, nsState := range current.RecordsByNameserver {
+		if nsState.Status != statusOK ||
+			slices.Contains(nsState.FailedTypes, "A") ||
+			slices.Contains(nsState.FailedTypes, "AAAA") ||
+			slices.Contains(nsState.FailedTypes, "CNAME") {
+			continue
+		}
+
+		answered = true
+
+		if len(nsState.Records["A"]) > 0 || len(nsState.Records["AAAA"]) > 0 {
+			return
+		}
+
+		for _, target := range nsState.Records["CNAME"] {
+			targets[target] = true
+		}
+	}
+
+	if !answered {
+		current.CNAMEAddresses = prevAddresses
+
+		return
+	}
+
+	for target := range targets {
+		ips, err := w.resolver.ResolveIPAddresses(ctx, target)
+		if err != nil {
+			w.logFailedLookup(
+				ctx,
+				"failed to follow CNAME",
+				"hostname", hostname,
+				"target", target,
+				"error", err,
+			)
+
+			current.CNAMEAddresses = prevAddresses
+
+			return
+		}
+
+		current.CNAMEAddresses = append(current.CNAMEAddresses, ips...)
+	}
+
+	// Still the empty list when every chain ends in no address.
+	slices.Sort(current.CNAMEAddresses)
+	current.CNAMEAddresses = slices.Compact(current.CNAMEAddresses)
+}
+
 // buildHostnameState saves each nameserver's response. A nameserver
-// that answered, even with NXDOMAIN or no records, is saved as ok; one
+// that answered, even with NXDOMAIN or no records, is saved as ok, with
-// that timed out or failed is saved as error with the reason, and its
+// the record types whose query failed; one that timed out or failed is
-// empty record set is not an answer.
+// saved as error with the reason, and its empty record set is not an
+// answer. prev is the hostname's state from the previous check, or nil.
 func buildHostnameState(
 	results map[string]*resolver.NameserverResponse,
+	prev *state.HostnameState,
 	now time.Time,
 ) *state.HostnameState {
 	hs := &state.HostnameState{
@@ -429,7 +592,7 @@ func buildHostnameState(
 
 	for ns, resp := range results {
 		nsState := &state.NameserverRecordState{
-			Records:     resp.Records,
+			Records:     maps.Clone(resp.Records),
 			Status:      statusOK,
 			LastChecked: now,
 		}
@@ -438,6 +601,15 @@ func buildHostnameState(
 			resp.Status == resolver.StatusError {
 			nsState.Status = statusError
 			nsState.Error = resp.Error
+		} else {
+			nsState.FailedTypes = resp.FailedTypes
+
+			var prevNS *state.NameserverRecordState
+			if prev != nil {
+				prevNS = prev.RecordsByNameserver[ns]
+			}
+
+			keepFailedTypes(nsState, prevNS)
 		}
 
 		hs.RecordsByNameserver[ns] = nsState
@@ -446,6 +618,27 @@ func buildHostnameState(
 	return hs
 }
 
+// keepFailedTypes copies into nsState, for each record type in its
+// FailedTypes, the records prevNS, the nameserver's state from the
+// previous check, holds for that type, which may be none. When prevNS
+// does not know them either, because the nameserver was new or failing
+// then or the type was in its UnknownTypes, the type goes in
+// nsState.UnknownTypes instead.
+func keepFailedTypes(nsState, prevNS *state.NameserverRecordState) {
+	for _, rtype := range nsState.FailedTypes {
+		if prevNS == nil || prevNS.Status != statusOK ||
+			slices.Contains(prevNS.UnknownTypes, rtype) {
+			nsState.UnknownTypes = append(nsState.UnknownTypes, rtype)
+
+			continue
+		}
+
+		if records, ok := prevNS.Records[rtype]; ok {
+			nsState.Records[rtype] = records
+		}
+	}
+}
+
 func (w *Watcher) detectHostnameChanges(
 	ctx context.Context,
 	hostname string,
@@ -455,11 +648,89 @@ func (w *Watcher) detectHostnameChanges(
 	w.detectNSDisappearances(ctx, hostname, prev, current)
 	w.detectNSFailures(ctx, hostname, prev, current)
 	w.detectInconsistencies(ctx, hostname, prev, current)
+	w.detectCNAMEAddressChanges(ctx, hostname, prev, current)
+}
+
+// isDomain reports whether name is a configured apex domain, whose own
+// records are checked and saved as a hostname's are.
+func (w *Watcher) isDomain(name string) bool {
+	return slices.Contains(w.config.Domains, name)
+}
+
+// nameLine is the line a notification about name's records starts with:
+// "Domain: " and the name for a configured apex domain, and
+// "Hostname: " otherwise.
+func (w *Watcher) nameLine(name string) string {
+	if w.isDomain(name) {
+		return "Domain: " + name
+	}
+
+	return "Hostname: " + name
+}
+
+// portNameLines lists the names that resolve to a port's address, the
+// configured apex domains on one line and the hostnames on the next,
+// leaving out a line that would name nothing.
+func (w *Watcher) portNameLines(names []string) string {
+	var domains, hostnames []string
+
+	for _, name := range names {
+		if w.isDomain(name) {
+			domains = append(domains, name)
+		} else {
+			hostnames = append(hostnames, name)
+		}
+	}
+
+	var lines []string
+
+	if len(domains) > 0 {
+		lines = append(lines, "Domains: "+strings.Join(domains, ", "))
+	}
+
+	if len(hostnames) > 0 {
+		lines = append(lines, "Hostnames: "+strings.Join(hostnames, ", "))
+	}
+
+	return strings.Join(lines, "\n")
+}
+
+// detectCNAMEAddressChanges notifies when the addresses at the end of
+// hostname's CNAME chain differ from those the previous check saved,
+// including a change from or to none. When the previous addresses are
+// not known (nil), as on the first check after loading a state file
+// written before they were saved, nothing is compared.
+func (w *Watcher) detectCNAMEAddressChanges(
+	ctx context.Context,
+	hostname string,
+	prev, current *state.HostnameState,
+) {
+	old, cur := prev.CNAMEAddresses, current.CNAMEAddresses
+	if old == nil || sliceEqual(old, cur) {
+		return
+	}
+
+	msg := fmt.Sprintf(
+		"%s\nOld: %s\nNew: %s",
+		w.nameLine(hostname),
+		strings.Join(old, ", "),
+		strings.Join(cur, ", "),
+	)
+
+	w.notify.SendNotification(
+		ctx,
+		"CNAME Address Change: "+hostname,
+		msg,
+		"warning",
+	)
 }
 
 // detectRecordChanges compares each nameserver's records with those of
 // the previous check. Only answers are compared: a nameserver that
-// failed on either check has no records to compare.
+// failed on either check has no records to compare. The records kept
+// for a record type whose query failed are compared too, but not those
+// of a type in UnknownTypes on either check, which the message leaves
+// out as well.
 func (w *Watcher) detectRecordChanges(
 	ctx context.Context,
 	hostname string,
@@ -471,15 +742,21 @@ func (w *Watcher) detectRecordChanges(
 			continue
 		}
 
-		if recordsEqual(prevNS.Records, cur.Records) {
+		unknown := slices.Concat(prevNS.UnknownTypes, cur.UnknownTypes)
+		oldRecords := withoutTypes(prevNS.Records, unknown)
+		newRecords := withoutTypes(cur.Records, unknown)
+
+		if recordsEqual(oldRecords, newRecords) {
 			continue
 		}
 
 		msg := fmt.Sprintf(
-			"Hostname: %s\nNameserver: %s\n"+
+			"%s\nNameserver: %s\n%s",
-				"Old: %v\nNew: %v",
+			w.nameLine(hostname), ns,
-			hostname, ns,
+			recordDifferences(
-			prevNS.Records, cur.Records,
+				"Old", oldRecords,
+				"New", newRecords,
+			),
 		)
 
 		w.notify.SendNotification(
@@ -502,8 +779,8 @@ func (w *Watcher) detectNSDisappearances(
 		}
 
 		msg := fmt.Sprintf(
-			"Hostname: %s\nNameserver: %s disappeared",
+			"%s\nNameserver: %s disappeared",
-			hostname, ns,
+			w.nameLine(hostname), ns,
 		)
 
 		w.notify.SendNotification(
@@ -532,8 +809,8 @@ func (w *Watcher) detectNSFailures(
 		switch {
 		case prevNS.Status == statusOK && cur.Status == statusError:
 			msg := fmt.Sprintf(
-				"Hostname: %s\nNameserver: %s\nError: %s",
+				"%s\nNameserver: %s\nError: %s",
-				hostname, ns, cur.Error,
+				w.nameLine(hostname), ns, cur.Error,
 			)
 
 			w.notify.SendNotification(
@@ -544,8 +821,8 @@ func (w *Watcher) detectNSFailures(
 			)
 		case prevNS.Status == statusError && cur.Status == statusOK:
 			msg := fmt.Sprintf(
-				"Hostname: %s\nNameserver: %s recovered",
+				"%s\nNameserver: %s recovered",
-				hostname, ns,
+				w.nameLine(hostname), ns,
 			)
 
 			w.notify.SendNotification(
@@ -565,12 +842,20 @@ func (w *Watcher) detectInconsistencies(
 ) {
 	for _, pair := range newlyDisagreeingPairs(prev, current) {
 		ns1, ns2 := pair[0], pair[1]
+		state1 := current.RecordsByNameserver[ns1]
+		state2 := current.RecordsByNameserver[ns2]
+
+		// The record types left out of the comparison are left out of
+		// the message too.
+		failed := slices.Concat(state1.FailedTypes, state2.FailedTypes)
 
 		msg := fmt.Sprintf(
-			"Hostname: %s\n%s: %v\n%s: %v",
+			"%s\n%s",
-			hostname,
+			w.nameLine(hostname),
-			ns1, current.RecordsByNameserver[ns1].Records,
+			recordDifferences(
-			ns2, current.RecordsByNameserver[ns2].Records,
+				ns1, withoutTypes(state1.Records, failed),
+				ns2, withoutTypes(state2.Records, failed),
+			),
 		)
 
 		w.notify.SendNotification(
@@ -587,7 +872,9 @@ func (w *Watcher) detectInconsistencies(
 // except pairs where both nameservers answered in prev and already
 // differed there. A nameserver missing from prev, or that failed there,
 // is paired with every nameserver it differs from. A nameserver that
-// failed in current has no records to compare and is in no pair.
+// failed in current has no records to compare and is in no pair. In
+// both checks, a record type whose query failed at either nameserver is
+// not compared.
 func newlyDisagreeingPairs(
 	prev, current *state.HostnameState,
 ) [][2]string {
@@ -604,9 +891,9 @@ func newlyDisagreeingPairs(
 
 	for i, ns1 := range nameservers {
 		for _, ns2 := range nameservers[i+1:] {
-			if recordsEqual(
+			if nameserversAgree(
-				current.RecordsByNameserver[ns1].Records,
+				current.RecordsByNameserver[ns1],
-				current.RecordsByNameserver[ns2].Records,
+				current.RecordsByNameserver[ns2],
 			) {
 				continue
 			}
@@ -616,7 +903,7 @@ func newlyDisagreeingPairs(
 
 			if ok1 && ok2 &&
 				prev1.Status == statusOK && prev2.Status == statusOK &&
-				!recordsEqual(prev1.Records, prev2.Records) {
+				!nameserversAgree(prev1, prev2) {
 				continue
 			}
 
@@ -644,8 +931,10 @@ func (w *Watcher) checkAllPorts(ctx context.Context) {
 	}
 
 	// Phase 3: Remove port state entries that no longer have
-	// any hostname referencing them.
+	// any hostname referencing them, and certificate entries for
+	// an address their name no longer has.
 	w.cleanupStalePorts(associations)
+	w.cleanupStaleCertificates()
 }
 
 // buildPortAssociations constructs a map from IP:port keys to
@@ -709,17 +998,85 @@ func parsePortKey(key string) (string, int) {
 }
 
 // cleanupStalePorts removes port state entries that are no
-// longer referenced by any hostname in the current DNS data.
+// longer referenced by any hostname in the current DNS data. An
+// entry saved for a configured name none of whose nameservers
+// answered is kept: that name's addresses are not known, not gone.
 func (w *Watcher) cleanupStalePorts(
 	currentAssociations map[string][]string,
 ) {
 	for _, key := range w.state.GetAllPortKeys() {
-		if _, exists := currentAssociations[key]; !exists {
+		if _, exists := currentAssociations[key]; exists {
-			w.state.DeletePortState(key)
+			continue
 		}
+
+		ps, ok := w.state.GetPortState(key)
+		if ok && slices.ContainsFunc(ps.Hostnames, w.noNameserverAnswered) {
+			continue
+		}
+
+		w.state.DeletePortState(key)
 	}
 }
 
+// cleanupStaleCertificates removes the certificate entries for an
+// address their name no longer resolves to. An entry saved for a name
+// none of whose nameservers answered is kept: that name's addresses are
+// not known, not gone.
+func (w *Watcher) cleanupStaleCertificates() {
+	for _, key := range w.state.GetAllCertificateKeys() {
+		ip, hostname := parseCertKey(key)
+
+		if slices.Contains(w.collectIPs(hostname), ip) ||
+			w.noNameserverAnswered(hostname) {
+			continue
+		}
+
+		w.state.DeleteCertificateState(key)
+	}
+}
+
+// parseCertKey splits an "ip:port:hostname" certificate key into its
+// address and hostname.
+func parseCertKey(key string) (string, string) {
+	lastColon := strings.LastIndex(key, ":")
+	if lastColon < 0 {
+		return "", key
+	}
+
+	ip, _ := parsePortKey(key[:lastColon])
+
+	return ip, key[lastColon+1:]
+}
+
+// isConfigured reports whether name is a configured domain or hostname.
+func (w *Watcher) isConfigured(name string) bool {
+	return w.isDomain(name) || slices.Contains(w.config.Hostnames, name)
+}
+
+// noNameserverAnswered reports whether name is a configured domain or
+// hostname and none of its nameservers answered on its last check.
+func (w *Watcher) noNameserverAnswered(name string) bool {
+	if !w.isConfigured(name) {
+		return false
+	}
+
+	hs, ok := w.state.GetHostnameState(name)
+	if !ok {
+		return false
+	}
+
+	for _, nsState := range hs.RecordsByNameserver {
+		if nsState.Status == statusOK {
+			return false
+		}
+	}
+
+	return true
+}
+
+// collectIPs returns the addresses saved for hostname: those in its
+// nameservers' A and AAAA records, and those at the end of its CNAME
+// chain.
 func (w *Watcher) collectIPs(hostname string) []string {
 	hs, ok := w.state.GetHostnameState(hostname)
 	if !ok {
@@ -738,6 +1095,10 @@ func (w *Watcher) collectIPs(hostname string) []string {
 		}
 	}
 
+	for _, ip := range hs.CNAMEAddresses {
+		ipSet[ip] = true
+	}
+
 	result := make([]string, 0, len(ipSet))
 	for ip := range ipSet {
 		result = append(result, ip)
@@ -784,8 +1145,8 @@ func (w *Watcher) checkSinglePort(
 		}
 
 		msg := fmt.Sprintf(
-			"Hosts: %s\nAddress: %s\nPort now %s",
+			"%s\nAddress: %s\nPort now %s",
-			strings.Join(hostnames, ", "), key, stateStr,
+			w.portNameLines(hostnames), key, stateStr,
 		)
 
 		w.notify.SendNotification(
@@ -796,9 +1157,24 @@ func (w *Watcher) checkSinglePort(
 		)
 	}
 
+	// A configured name on the saved list none of whose nameservers
+	// answered stays on it, so the entry is kept when the other names
+	// stop resolving to this address.
+	savedHostnames := slices.Clone(hostnames)
+
+	if hasPrev {
+		for _, name := range prev.Hostnames {
+			if !slices.Contains(hostnames, name) && w.noNameserverAnswered(name) {
+				savedHostnames = append(savedHostnames, name)
+			}
+		}
+
+		sort.Strings(savedHostnames)
+	}
+
 	w.state.SetPortState(key, &state.PortState{
 		Open:        result.Open,
-		Hostnames:   hostnames,
+		Hostnames:   savedHostnames,
 		LastChecked: now,
 	})
 }
@@ -982,22 +1358,6 @@ func (w *Watcher) checkTLSExpiry(
 		return
 	}
 
-	// Deduplicate expiry warnings: don't re-notify for the same
-	// hostname within the TLS check interval.
-	dedupKey := fmt.Sprintf("expiry:%s:%s", hostname, ip)
-
-	w.expiryNotifiedMu.Lock()
-
-	lastNotified, seen := w.expiryNotified[dedupKey]
-	if seen && time.Since(lastNotified) < w.config.TLSInterval {
-		w.expiryNotifiedMu.Unlock()
-
-		return
-	}
-
-	w.expiryNotified[dedupKey] = time.Now()
-	w.expiryNotifiedMu.Unlock()
-
 	msg := fmt.Sprintf(
 		"Host: %s\nIP: %s\nCN: %s\n"+
 			"Expires: %s (%.0f days)",
@@ -1023,8 +1383,11 @@ func (w *Watcher) saveState() {
 
 // maybeSendTestNotification sends a startup status notification
 // after the first full scan completes, if SEND_TEST_NOTIFICATION
-// is enabled. The message is clearly informational ("all ok")
+// is enabled. The message is informational, not an error or anomaly
-// and not an error or anomaly alert.
+// alert. It is written before it reaches any endpoint, so it claims
+// nothing about whether the endpoints work. Domains and hostnames are
+// counted from the configuration: the state's hostnames also hold each
+// apex domain's own records.
 func (w *Watcher) maybeSendTestNotification(ctx context.Context) {
 	if !w.config.SendTestNotification {
 		return
@@ -1036,9 +1399,10 @@ func (w *Watcher) maybeSendTestNotification(ctx context.Context) {
 		"dnswatcher has started and completed its initial scan.\n"+
 			"Monitoring %d domain(s) and %d hostname(s).\n"+
 			"Tracking %d port endpoint(s) and %d TLS certificate(s).\n"+
-			"All notification channels are working.",
+			"This is a test notification, sent to every configured "+
-		len(snap.Domains),
+			"notification endpoint.",
-		len(snap.Hostnames),
+		len(w.config.Domains),
+		len(w.config.Hostnames),
 		len(snap.Ports),
 		len(snap.Certificates),
 	)
@@ -1064,6 +1428,33 @@ func toSet(items []string) map[string]bool {
 	return set
 }
 
+// nameserversAgree reports whether two nameservers' states from the same
+// check hold the same records, leaving out the record types either lists
+// in FailedTypes: the records held for those are kept from an earlier
+// check, or not known.
+func nameserversAgree(a, b *state.NameserverRecordState) bool {
+	failed := slices.Concat(a.FailedTypes, b.FailedTypes)
+
+	return recordsEqual(
+		withoutTypes(a.Records, failed), withoutTypes(b.Records, failed),
+	)
+}
+
+// withoutTypes returns a copy of records without the record types in
+// types.
+func withoutTypes(
+	records map[string][]string,
+	types []string,
+) map[string][]string {
+	records = maps.Clone(records)
+
+	for _, rtype := range types {
+		delete(records, rtype)
+	}
+
+	return records
+}
+
 func recordsEqual(
 	a, b map[string][]string,
 ) bool {
@@ -1081,6 +1472,54 @@ func recordsEqual(
 	return true
 }
 
+// recordDifferences describes, in sorted order of type, each record
+// type whose values differ between a and b: a line naming the type,
+// then a line with a's values after labelA and one with b's after
+// labelB. Types with the same values in both are left out.
+func recordDifferences(
+	labelA string, a map[string][]string,
+	labelB string, b map[string][]string,
+) string {
+	types := make([]string, 0, len(a)+len(b))
+
+	for recordType := range a {
+		types = append(types, recordType)
+	}
+
+	for recordType := range b {
+		if _, ok := a[recordType]; !ok {
+			types = append(types, recordType)
+		}
+	}
+
+	sort.Strings(types)
+
+	var lines []string
+
+	for _, recordType := range types {
+		if sliceEqual(a[recordType], b[recordType]) {
+			continue
+		}
+
+		lines = append(lines,
+			"Type: "+recordType,
+			labelA+": "+joinValues(a[recordType]),
+			labelB+": "+joinValues(b[recordType]),
+		)
+	}
+
+	return strings.Join(lines, "\n")
+}
+
+// joinValues lists record values separated by commas, or says none.
+func joinValues(values []string) string {
+	if len(values) == 0 {
+		return "none"
+	}
+
+	return strings.Join(values, ", ")
+}
+
 func sliceEqual(a, b []string) bool {
 	if len(a) != len(b) {
 		return false

internal/watcher/watcher_test.go

@@ -26,18 +26,22 @@ import (
 
 // The watcher looks these names up in live DNS with the real resolver,
 // so tests assert on what the watcher does with the answers, never on
-// the records these zones publish. testHost's nameservers and addresses
+// the records these zones publish. The nameservers of testHost and
-// stay the same from one check to the next, which the tests that check
+// testSmallDomain stay the same between a test looking them up and its
-// it twice rely on, and testSmallDomain's nameservers stay the same
+// check. Every query a check sends is one more that can be lost, so the
-// between a test looking them up and its check. A domain check looks up
+// tests keep them few. A check asks each of a name's nameservers about
-// each nameserver's addresses, about a second per nameserver, so the
+// every record type, and both names have two. A domain check also looks
-// tests that check a domain use testSmallDomain, which has two
+// up each nameserver's addresses at every nameserver of the zone that
-// nameservers, and check it once. The tests that query testDomain's
+// nameserver is in: testSmallDomain's nameservers are in zones with two
-// nameservers directly do no domain check.
+// nameservers, while a domain whose nameservers are in, say,
+// cloudflare.com, which has five, makes each domain check much longer.
+// A test checks a domain only when it is about domains, and checks once,
+// from saved state it builds, rather than twice. The tests that query
+// testDomain's nameservers directly do no domain check.
 const (
 	testDomain      = "google.com"
-	testSmallDomain = "example.com"
+	testSmallDomain = "desec.io"
-	testHost        = "cloudflare.com"
+	testHost        = "example.org"
 	testIssuer      = "DigiCert"
 )
 
@@ -259,55 +263,48 @@ func checkOnce(
 
 // runChecks builds a watcher, lets prepare set up the saved state and
 // stand-ins it starts from, and runs its checks once against live DNS.
-// If change is not nil, change then alters the saved state or stand-ins
+// When the check finds no fresh address for a name (see checkOnce), the
-// and the checks run a second time. When either check finds no fresh
+// watcher is thrown away and all of this runs again on a new one, so a
-// address for a name (see checkOnce), the watcher is thrown away and
+// failed attempt leaves nothing behind in the saved state, the
-// all of this runs again on a new one, so a failed attempt leaves
+// stand-ins or the notifications.
-// nothing behind in the saved state, the stand-ins or the notifications.
 func runChecks(
 	t *testing.T,
 	cfg *config.Config,
-	prepare, change func(deps *testDeps),
+	prepare func(deps *testDeps),
-) *testDeps {
+) (*watcher.Watcher, *testDeps) {
 	t.Helper()
 
-	var deps *testDeps
+	var (
+		w    *watcher.Watcher
+		deps *testDeps
+	)
 
 	livednstest.Retry(t, "watcher checks", func(ctx context.Context) error {
-		var w *watcher.Watcher
-
 		w, deps = newTestWatcher(t, cfg)
 
 		if prepare != nil {
 			prepare(deps)
 		}
 
-		err := checkOnce(ctx, w, deps)
-		if err != nil || change == nil {
-			return err
-		}
-
-		change(deps)
-
 		return checkOnce(ctx, w, deps)
 	})
 
-	return deps
+	return w, deps
 }
 
-// lookupNameservers returns the nameservers live DNS lists for domain,
+// lookupNameservers returns the nameservers live DNS lists for name,
 // for a test to save in the state its check starts from.
-func lookupNameservers(t *testing.T, domain string) []string {
+func lookupNameservers(t *testing.T, name string) []string {
 	t.Helper()
 
 	res := resolver.NewFromLogger(slog.Default())
 
 	var nameservers []string
 
-	livednstest.Retry(t, "LookupNS("+domain+")", func(ctx context.Context) error {
+	livednstest.Retry(t, "LookupNS("+name+")", func(ctx context.Context) error {
 		var err error
 
-		nameservers, err = res.LookupNS(ctx, domain)
+		nameservers, err = res.LookupNS(ctx, name)
 
 		return err
 	})
@@ -315,7 +312,8 @@ func lookupNameservers(t *testing.T, domain string) []string {
 	return nameservers
 }
 
-// addresses returns the A and AAAA values saved for a hostname.
+// addresses returns the A and AAAA values saved for a hostname, and the
+// addresses saved at the end of its CNAME chain.
 func addresses(hs *state.HostnameState) []string {
 	var ips []string
 
@@ -324,7 +322,7 @@ func addresses(hs *state.HostnameState) []string {
 		ips = append(ips, nsState.Records["AAAA"]...)
 	}
 
-	return ips
+	return append(ips, hs.CNAMEAddresses...)
 }
 
 // assertNotified checks that a notification with this title and
@@ -370,10 +368,18 @@ func TestFirstRunBaseline(t *testing.T) {
 	cfg.Domains = []string{testSmallDomain}
 	cfg.Hostnames = []string{testHost}
 
-	deps := runChecks(t, cfg, nil, nil)
+	_, deps := runChecks(t, cfg, nil)
 
 	assertNoNotifications(t, deps)
 	assertStatePopulated(t, deps)
+
+	// testHost answers with an address, so the check saves an empty list
+	// of CNAME addresses for it; nil would mean the check did not look
+	// at whether to follow a CNAME.
+	hs, _ := deps.state.GetHostnameState(testHost)
+	if hs.CNAMEAddresses == nil || len(hs.CNAMEAddresses) != 0 {
+		t.Errorf("saved CNAME addresses %#v, want []", hs.CNAMEAddresses)
+	}
 }
 
 func assertNoNotifications(
@@ -422,7 +428,7 @@ func TestDomainPortAndTLSChecks(t *testing.T) {
 	cfg := defaultTestConfig(t)
 	cfg.Domains = []string{testSmallDomain}
 
-	deps := runChecks(t, cfg, nil, nil)
+	_, deps := runChecks(t, cfg, nil)
 
 	snap := deps.state.GetSnapshot()
 
@@ -462,11 +468,11 @@ func TestNSChangeDetection(t *testing.T) {
 	cfg.Domains = []string{testSmallDomain}
 
 	// The saved state lists nameservers that live DNS does not.
-	deps := runChecks(t, cfg, func(deps *testDeps) {
+	_, deps := runChecks(t, cfg, func(deps *testDeps) {
 		deps.state.SetDomainState(testSmallDomain, &state.DomainState{
 			Nameservers: []string{oldNS1, oldNS2},
 		})
-	}, nil)
+	})
 
 	assertNotified(t, deps, "NS Change: "+testSmallDomain, "warning")
 
@@ -476,6 +482,119 @@ func TestNSChangeDetection(t *testing.T) {
 	}
 }
 
+// TestDomainThatDoesNotExist checks a .com domain that does not exist,
+// with nameservers and records saved by an earlier check. The .com
+// servers answer that it does not exist, so it is saved with nxdomain
+// set and no nameservers, an NS Change removes them all, and its saved
+// records are removed rather than asked for at the .com servers.
+func TestDomainThatDoesNotExist(t *testing.T) {
+	t.Parallel()
+
+	const domain = "dnswatcher-test-does-not-exist.com"
+
+	cfg := defaultTestConfig(t)
+	cfg.Domains = []string{domain}
+
+	var deps *testDeps
+
+	livednstest.Retry(t, "watcher checks", func(ctx context.Context) error {
+		var w *watcher.Watcher
+
+		w, deps = newTestWatcher(t, cfg)
+
+		deps.state.SetDomainState(domain, &state.DomainState{
+			Nameservers: []string{oldNS1, oldNS2},
+		})
+		deps.state.SetHostnameState(domain, &state.HostnameState{
+			RecordsByNameserver: map[string]*state.NameserverRecordState{
+				oldNS1: {
+					Records: map[string][]string{"A": {oldIP}},
+					Status:  "ok",
+				},
+			},
+		})
+
+		started := time.Now()
+
+		w.RunOnce(ctx)
+
+		// When no server answered, the domain's state is not saved.
+		ds, _ := deps.state.GetDomainState(domain)
+		if ds.LastChecked.Before(started) {
+			return fmt.Errorf("%s: %w", domain, livednstest.ErrNoAnswer)
+		}
+
+		return nil
+	})
+
+	ds, _ := deps.state.GetDomainState(domain)
+	if !ds.NXDomain || len(ds.Nameservers) != 0 {
+		t.Errorf("saved nxdomain %v and nameservers %v, want true and none",
+			ds.NXDomain, ds.Nameservers)
+	}
+
+	if hs, ok := deps.state.GetHostnameState(domain); ok {
+		t.Errorf("records saved for %s: %v", domain, hs.RecordsByNameserver)
+	}
+
+	assertNotified(t, deps, "NS Change: "+domain, "warning")
+
+	// That is the only notification, and it removes both nameservers,
+	// in either order.
+	for _, n := range deps.notifier.getNotifications() {
+		removed := strings.TrimPrefix(
+			n.Message, "Domain: "+domain+"\nAdded: \nRemoved: ",
+		)
+		if removed != oldNS1+", "+oldNS2 && removed != oldNS2+", "+oldNS1 {
+			t.Errorf("unexpected notification: %v", n)
+		}
+	}
+}
+
+// TestDomainWithNoDelegationOfItsOwn checks a domain with no delegation
+// of its own: codeberg.page is on the public suffix list, so
+// docs.codeberg.page is a domain, but the .page servers delegate only
+// codeberg.page, whose servers answer for it. It is saved with no
+// nameservers and without nxdomain, and its records, asked at the
+// codeberg.page servers, are saved. Those are testSmallDomain's two
+// nameservers; github.io, the zone of the README's example, has eight.
+func TestDomainWithNoDelegationOfItsOwn(t *testing.T) {
+	t.Parallel()
+
+	const domain = "docs.codeberg.page"
+
+	cfg := defaultTestConfig(t)
+	cfg.Domains = []string{domain}
+
+	var deps *testDeps
+
+	livednstest.Retry(t, "watcher checks", func(ctx context.Context) error {
+		var w *watcher.Watcher
+
+		w, deps = newTestWatcher(t, cfg)
+
+		err := checkOnce(ctx, w, deps)
+
+		// A domain saved as not existing has no records to wait for;
+		// the checks below fail on it.
+		if ds, ok := deps.state.GetDomainState(domain); ok && ds.NXDomain {
+			return nil
+		}
+
+		return err
+	})
+
+	ds, _ := deps.state.GetDomainState(domain)
+	if ds.NXDomain || len(ds.Nameservers) != 0 {
+		t.Errorf("saved nxdomain %v and nameservers %v, want false and none",
+			ds.NXDomain, ds.Nameservers)
+	}
+
+	if _, ok := deps.state.GetHostnameState(domain); !ok {
+		t.Errorf("no records saved for %s", domain)
+	}
+}
+
 func TestNSAddressChangeDetection(t *testing.T) {
 	t.Parallel()
 
@@ -486,7 +605,7 @@ func TestNSAddressChangeDetection(t *testing.T) {
 
 	// The saved state lists the nameservers live DNS lists, each at an
 	// address live DNS never returns.
-	deps := runChecks(t, cfg, func(deps *testDeps) {
+	_, deps := runChecks(t, cfg, func(deps *testDeps) {
 		nsAddresses := make(map[string][]string, len(nameservers))
 		for _, ns := range nameservers {
 			nsAddresses[ns] = []string{oldIP}
@@ -496,7 +615,7 @@ func TestNSAddressChangeDetection(t *testing.T) {
 			Nameservers:         nameservers,
 			NameserverAddresses: nsAddresses,
 		})
-	}, nil)
+	})
 
 	title := "NS Address Change: " + testSmallDomain
 	ds, _ := deps.state.GetDomainState(testSmallDomain)
@@ -542,12 +661,12 @@ func TestNSAddedAndRemovedIsNoAddressChange(t *testing.T) {
 	// The saved state lists oldNS1, which live DNS does not, in place of
 	// the first nameserver live DNS lists, so that the check finds that
 	// one added and oldNS1 removed. Only oldNS1 has addresses saved.
-	deps := runChecks(t, cfg, func(deps *testDeps) {
+	_, deps := runChecks(t, cfg, func(deps *testDeps) {
 		deps.state.SetDomainState(testSmallDomain, &state.DomainState{
 			Nameservers:         append([]string{oldNS1}, nameservers[1:]...),
 			NameserverAddresses: map[string][]string{oldNS1: {oldIP}},
 		})
-	}, nil)
+	})
 
 	if n := countNotifications(deps, "NS Change: "+testSmallDomain); n != 1 {
 		t.Errorf("sent %d NS changes, want 1", n)
@@ -565,15 +684,17 @@ func TestRecordChangeDetection(t *testing.T) {
 	cfg := defaultTestConfig(t)
 	cfg.Hostnames = []string{testHost}
 
-	// Between the checks, save for every nameserver an address live DNS
+	nameservers := lookupNameservers(t, testHost)
-	// never returns.
+
-	deps := runChecks(t, cfg, nil, func(deps *testDeps) {
+	// The saved state has every nameserver live DNS lists answering
-		hs, _ := deps.state.GetHostnameState(testHost)
+	// with an address live DNS never returns.
-		for _, nsState := range hs.RecordsByNameserver {
+	_, deps := runChecks(t, cfg, func(deps *testDeps) {
-			nsState.Records = map[string][]string{"A": {oldIP}}
+		byNameserver := make(map[string]*state.NameserverRecordState)
+		for _, ns := range nameservers {
+			byNameserver[ns] = answered(map[string][]string{"A": {oldIP}})
 		}
 
-		deps.state.SetHostnameState(testHost, hs)
+		deps.state.SetHostnameState(testHost, saved(byNameserver))
 	})
 
 	assertNotified(t, deps, "Record Change: "+testHost, "warning")
@@ -585,12 +706,15 @@ func TestPortStateChange(t *testing.T) {
 	cfg := defaultTestConfig(t)
 	cfg.Hostnames = []string{testHost}
 
-	// Between the checks, every port closes.
+	w, deps := runChecks(t, cfg, nil)
-	deps := runChecks(t, cfg, nil, func(deps *testDeps) {
+
+	// Every port closes, and the port checks run again. They look
+	// nothing up.
 	deps.portChecker.mu.Lock()
 	deps.portChecker.closed = true
 	deps.portChecker.mu.Unlock()
-	})
+
+	w.CheckAllPorts(t.Context())
 
 	hs, _ := deps.state.GetHostnameState(testHost)
 	assertNotified(
@@ -610,39 +734,60 @@ func TestTLSExpiryWarning(t *testing.T) {
 	cfg := defaultTestConfig(t)
 	cfg.Hostnames = []string{testHost}
 
-	deps := runChecks(t, cfg, expiresInThreeDays, nil)
+	_, deps := runChecks(t, cfg, expiresInThreeDays)
 
 	assertNotified(t, deps, "TLS Expiry Warning: "+testHost, "warning")
 }
 
-func TestTLSExpiryWarningDedup(t *testing.T) {
+// TestTLSExpiryWarningEachCheck runs the TLS checks three times in a
+// row on hostname and port state built here, for a certificate that
+// expires within the warning period. Each check warns once, whether the
+// TLS interval is a nanosecond, shorter than the time between two
+// checks, or a day, longer than it.
+func TestTLSExpiryWarningEachCheck(t *testing.T) {
+	t.Parallel()
+
+	title := "TLS Expiry Warning: " + host
+
+	for _, interval := range []time.Duration{time.Nanosecond, 24 * time.Hour} {
+		t.Run(interval.String(), func(t *testing.T) {
 			t.Parallel()
 
 			cfg := defaultTestConfig(t)
-	cfg.Hostnames = []string{testHost}
+			cfg.Hostnames = []string{host}
-	cfg.TLSInterval = 24 * time.Hour
+			cfg.TLSInterval = interval
 
-	title := "TLS Expiry Warning: " + testHost
+			// The TLS checks read the saved hostname and port state and
+			// look nothing up, so the watcher has no resolver.
+			deps := newTestDeps(t, cfg)
+			w := watcher.NewForTest(
+				cfg, deps.state, nil,
+				deps.portChecker, deps.tlsChecker, deps.notifier,
+			)
 
-	// The second check comes within the TLS interval of the first,
+			expiresInThreeDays(deps)
-	// so it must not warn again.
+			deps.state.SetHostnameState(host, saved(
-	var warnings int
+				map[string]*state.NameserverRecordState{
-
+					nsA: answered(map[string][]string{"A": {ip1}}),
-	deps := runChecks(t, cfg, expiresInThreeDays, func(deps *testDeps) {
+				},
-		warnings = countNotifications(deps, title)
+			))
+			deps.state.SetPortState(ip1+":443", &state.PortState{
+				Open: true, Hostnames: []string{host},
 			})
 
-	if warnings == 0 {
+			for check := 1; check <= 3; check++ {
-		t.Fatal("expected expiry warnings from the first check")
+				w.RunTLSChecks(t.Context())
-	}
 
 				got := countNotifications(deps, title)
-	if got != warnings {
+				if got != check {
-		t.Errorf(
+					t.Fatalf(
-			"expected %d expiry warnings (dedup), got %d",
+						"after check %d: %d expiry warnings, want %d",
-			warnings, got,
+						check, got, check,
 					)
 				}
+			}
+		})
+	}
 }
 
 func TestGracefulShutdown(t *testing.T) {
@@ -761,7 +906,7 @@ func TestDNSRunsBeforePortAndTLSChecks(t *testing.T) {
 	cfg.Hostnames = []string{testHost}
 
 	// The saved state says the last check found testHost at oldIP.
-	deps := runChecks(t, cfg, func(deps *testDeps) {
+	_, deps := runChecks(t, cfg, func(deps *testDeps) {
 		deps.state.SetHostnameState(testHost, &state.HostnameState{
 			RecordsByNameserver: map[string]*state.NameserverRecordState{
 				oldNS1: {
@@ -770,7 +915,7 @@ func TestDNSRunsBeforePortAndTLSChecks(t *testing.T) {
 				},
 			},
 		})
-	}, nil)
+	})
 
 	snap := deps.state.GetSnapshot()
 
@@ -843,18 +988,27 @@ func TestSendTestNotification_ViaRun(t *testing.T) {
 
 	notifications := deps.notifier.getNotifications()
 
+	// No names are configured, so every count is 0.
+	wantMessage := "dnswatcher has started and completed its initial scan.\n" +
+		"Monitoring 0 domain(s) and 0 hostname(s).\n" +
+		"Tracking 0 port endpoint(s) and 0 TLS certificate(s).\n" +
+		"This is a test notification, sent to every configured " +
+		"notification endpoint."
+
 	found := false
 
 	for _, n := range notifications {
 		if n.Priority == "success" &&
-			n.Title == "✅ dnswatcher startup complete" {
+			n.Title == "✅ dnswatcher startup complete" &&
+			n.Message == wantMessage {
 			found = true
 		}
 	}
 
 	if !found {
 		t.Errorf(
-			"expected startup test notification, got: %v",
+			"expected startup test notification with message %q, got: %v",
+			wantMessage,
 			notifications,
 		)
 	}
@@ -901,21 +1055,20 @@ func TestNSFailureAndRecovery(t *testing.T) {
 	cfg := defaultTestConfig(t)
 	cfg.Hostnames = []string{testHost}
 
-	// Between the checks, save every nameserver the first check found
+	nameservers := lookupNameservers(t, testHost)
-	// as one that did not answer, and add, as answering, one that live
+
-	// DNS does not list, which then disappears.
+	// The saved state has every nameserver live DNS lists as one that
-	deps := runChecks(t, cfg, nil, func(deps *testDeps) {
+	// did not answer, and, as answering, one that live DNS does not
-		hs, _ := deps.state.GetHostnameState(testHost)
+	// list, which then disappears.
-		for ns := range hs.RecordsByNameserver {
+	_, deps := runChecks(t, cfg, func(deps *testDeps) {
-			hs.RecordsByNameserver[ns] = failed()
+		byNameserver := map[string]*state.NameserverRecordState{
+			oldNS1: answered(map[string][]string{"A": {oldIP}}),
+		}
+		for _, ns := range nameservers {
+			byNameserver[ns] = failed()
 		}
 
-		hs.RecordsByNameserver[oldNS1] = &state.NameserverRecordState{
+		deps.state.SetHostnameState(testHost, saved(byNameserver))
-			Records: map[string][]string{"A": {oldIP}},
-			Status:  "ok",
-		}
-
-		deps.state.SetHostnameState(testHost, hs)
 	})
 
 	assertNotified(t, deps, "NS Failure: "+testHost, "error")

package.json

@@ -0,0 +1,9 @@
+{
+    "name": "dnswatcher-tooling",
+    "version": "0.0.0",
+    "private": true,
+    "description": "Pins the prettier that script/fmt and script/fmt-check run against this repo's markdown. Not a JavaScript project; nothing here is imported, published, or shipped.",
+    "devDependencies": {
+        "prettier": "3.9.6"
+    }
+}

script/bootstrap

@@ -3,11 +3,12 @@
 # this repo. Idempotent: every install is guarded by a check so already
 # installed tools are skipped. Base tooling comes from nix, apt, brew,
 # or apk (detected in that order); assumes nothing is present.
-# goimports is not installed here: script/fmt and script/fmt-check run
+# goimports is not installed here: script/fmt and script/fmt-check-go
-# it with `go run` at a pinned commit.
+# run it with `go run` at a pinned commit.
 # The linter is NOT installed here: golangci-lint runs via docker only
 # (script/lint), pinned by image digest, so its only prerequisite is a
-# working docker.
+# working docker. Nor is prettier: script/fmt and
+# script/fmt-check-markdown run it in a container from Dockerfile.fmt.
 set -eu
 
 ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
@@ -66,11 +67,12 @@ main() {
     if missing make; then pkg_install gnumake make make make; fi
     if missing go; then pkg_install go golang go go; fi
 
-    # Linting runs via docker only (script/lint). Warn, don't fail:
+    # Linting and the markdown formatter run via docker only. Warn,
-    # everything except `make lint` works without it.
+    # don't fail: building and testing work without it.
     if missing docker; then
         echo "bootstrap: WARNING: docker not found; install it to" \
-            "run make lint and make docker." >&2
+            "run make lint, make fmt, make fmt-check, make check" \
+            "and make docker." >&2
     fi
 
     go mod download

script/cibuild

@@ -1,18 +1,23 @@
 #!/bin/sh
 # script/cibuild: run the CI build. The Dockerfile's lint stage runs
-# make fmt-check and golangci-lint; its builder stage runs make test
+# the Go half of make fmt-check and golangci-lint; its builder stage
-# and make build.
+# runs make test and make build. The markdown half of make fmt-check
+# runs after that build, as its own build of Dockerfile.fmt, because
+# there is no docker inside a docker build.
 #
 # --no-cache-filter=lint,builder runs both stages on every invocation;
 # otherwise an unchanged tree is served from the layer cache and passes
-# without linting or querying live DNS.
+# without linting or querying live DNS. script/fmt-check-markdown busts
+# its own cache the same way.
 set -eu
 
-ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd -P)"
+ROOT="$(cd "$SCRIPT_DIR/.." && pwd -P)"
 
 main() {
     cd "$ROOT"
     docker build --no-cache-filter=lint,builder .
+    "$SCRIPT_DIR/fmt-check-markdown"
 }
 
 main "$@"

script/docker

@@ -14,8 +14,8 @@ main() {
     cd "$ROOT"
     # Own line: a failing command substitution inside an argument does
     # not trip `set -e`, so the inline form degrades silently to an
-    # empty constant. VERSION is computed here because .dockerignore
+    # empty constant. The VERSION build arg takes precedence over what
-    # excludes .git, so `git describe` in a build stage cannot find it.
+    # the build would derive from the .git in its context.
     version="$(git describe --tags --always --dirty 2>/dev/null || true)"
     [ -n "$version" ] || version="unknown"
     docker build --no-cache-filter=lint,builder \

script/fmt

@@ -1,19 +1,65 @@
 #!/bin/sh
-# script/fmt: format all files (writes).
+# script/fmt: format all files (writes). Go with gofmt and goimports on
+# the host, markdown with the prettier pinned by Dockerfile.fmt.
 #
 # goimports runs with `go run` at a pinned commit, never from PATH, so
 # every machine formats with the same version and nothing installs it.
+#
+# The markdown pass is a `docker build --output type=local` rather than a
+# `docker run -v`, so it needs no bind mount and behaves the same against
+# a remote daemon; the formatted documents come back out of the build and
+# are copied over the tree here.
+#
+# Unlike script/fmt-check-markdown this does not bust the cache: it is
+# not a gate, and any edit to a document changes the COPY layer above the
+# prettier step, so a cached result is a result over this exact tree.
 set -eu
 
 ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
 
-# goimports v0.42.0, 2026-08-07. Must match script/fmt-check.
+# goimports v0.42.0, 2026-08-07. Must match script/fmt-check-go.
 GOIMPORTS_REF="golang.org/x/tools/cmd/goimports@009367f5c17a8d4c45a961a3a509277190a9a6f0"
 
+# Must match the export stage name in Dockerfile.fmt.
+stage=fmt-out
+
+die() {
+    echo "script/fmt: $*" >&2
+    exit 1
+}
+
 main() {
     cd "$ROOT"
     gofmt -s -w .
     go run "$GOIMPORTS_REF" -w .
+
+    tmp="$(mktemp -d "${TMPDIR:-/tmp}/dnswatcher-fmt.XXXXXX")"
+    trap 'rm -rf "$tmp"' EXIT INT TERM
+
+    docker build \
+        --target "$stage" \
+        --output "type=local,dest=$tmp/out" \
+        -f Dockerfile.fmt .
+
+    # An empty export means prettier was handed nothing, which must not
+    # read as "already formatted".
+    (cd "$tmp/out" && find . -type f -name '*.md') |
+        sed 's|^\./||' | LC_ALL=C sort >"$tmp/files"
+
+    [ -s "$tmp/files" ] ||
+        die "the formatting build produced no markdown; the build" \
+            "context reached prettier empty"
+
+    # Copied only where the bytes differ, so an already-formatted tree
+    # keeps its timestamps and says nothing.
+    while IFS= read -r f; do
+        [ -n "$f" ] || continue
+        if [ -f "$f" ] && cmp -s "$tmp/out/$f" "$f"; then
+            continue
+        fi
+        cp "$tmp/out/$f" "$f"
+        echo "prettier: reformatted $f"
+    done <"$tmp/files"
 }
 
 main "$@"

script/fmt-check

@@ -1,27 +1,14 @@
 #!/bin/sh
 # script/fmt-check: check formatting (read-only). Same tools and scope
-# as script/fmt, but fails instead of writing.
+# as script/fmt, but fails instead of writing: the Go on the host, the
+# markdown with prettier in a container.
 set -eu
 
-ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd -P)"
-
-# goimports v0.42.0, 2026-08-07. Must match script/fmt.
-GOIMPORTS_REF="golang.org/x/tools/cmd/goimports@009367f5c17a8d4c45a961a3a509277190a9a6f0"
 
 main() {
-    cd "$ROOT"
+    "$SCRIPT_DIR/fmt-check-go"
-    files="$(gofmt -s -l .)"
+    "$SCRIPT_DIR/fmt-check-markdown"
-    if [ -n "$files" ]; then
-        echo "gofmt: files not formatted:" >&2
-        echo "$files" >&2
-        exit 1
-    fi
-    files="$(go run "$GOIMPORTS_REF" -l .)"
-    if [ -n "$files" ]; then
-        echo "goimports: files not formatted:" >&2
-        echo "$files" >&2
-        exit 1
-    fi
 }
 
 main "$@"

script/fmt-check-go

@@ -0,0 +1,31 @@
+#!/bin/sh
+# script/fmt-check-go: fail unless every Go source is formatted the way
+# script/fmt would leave it, and name the files that are not. Read-only.
+#
+# Its own script because the Dockerfile's lint stage runs this half
+# alone: there is no docker inside a docker build to run the markdown
+# half in.
+set -eu
+
+ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
+
+# goimports v0.42.0, 2026-08-07. Must match script/fmt.
+GOIMPORTS_REF="golang.org/x/tools/cmd/goimports@009367f5c17a8d4c45a961a3a509277190a9a6f0"
+
+main() {
+    cd "$ROOT"
+    files="$(gofmt -s -l .)"
+    if [ -n "$files" ]; then
+        echo "gofmt: files not formatted:" >&2
+        echo "$files" >&2
+        exit 1
+    fi
+    files="$(go run "$GOIMPORTS_REF" -l .)"
+    if [ -n "$files" ]; then
+        echo "goimports: files not formatted:" >&2
+        echo "$files" >&2
+        exit 1
+    fi
+}
+
+main "$@"

script/fmt-check-markdown

@@ -0,0 +1,29 @@
+#!/bin/sh
+# script/fmt-check-markdown: fail unless every .md is formatted the way
+# script/fmt would leave it. Read-only.
+#
+# prettier is never installed on the host: it runs in a container built
+# from Dockerfile.fmt, pinned by package.json and yarn.lock.
+# --no-cache-filter is here for the reason script/lint gives: a cached
+# build checks nothing.
+#
+# Its own script because script/cibuild runs this half alone, after the
+# Dockerfile's lint stage has checked the Go.
+set -eu
+
+ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
+
+# Must match the markdown check stage name in Dockerfile.fmt.
+stage=fmt-check
+
+main() {
+    cd "$ROOT"
+    docker build \
+        --progress=plain \
+        --no-cache-filter="$stage" \
+        --target "$stage" \
+        -f Dockerfile.fmt \
+        .
+}
+
+main "$@"

yarn.lock

@@ -0,0 +1,8 @@
+# THIS IS AN AUTOGENERATED FILE. DO NOT EDIT THIS FILE DIRECTLY.
+# yarn lockfile v1
+
+
+prettier@3.9.6:
+  version "3.9.6"
+  resolved "https://registry.yarnpkg.com/prettier/-/prettier-3.9.6.tgz#b3ea5146515d40fc53f18aa63f74dfab1e10dbf6"
+  integrity sha512-OpN0zzVdiaiAhxpuuj5efpIS4sY9j7bY6uR5mnj5yPzGkdkjNKSJeUThPb60Jw29QuAZgA4o+/iB49kFiaBX6g==