test: bound watcher live DNS and serialise packages after rebase

Rebasing this branch onto next surfaced two failures that the branch
did not have in isolation. Both come from the change this PR makes:
the watcher package now queries live DNS, and it is the second package
to do so.

Burst fan-out in the watcher package. All 13 watcher tests are
parallel and each runs a full iterative resolution, so they hit the
same root servers in the same instant and get rate-limited. This is
exactly the pathology internal/resolver/livedns_test.go was written to
prevent (9cb2c2b); that gate is package-scoped and cannot reach into
this package's test binary, so liveWatcherGate is its counterpart
here, acquired in newTestWatcher and released when the test ends.

Cross-package oversubscription. Go runs package binaries in parallel,
so with both live-DNS packages in flight their gates sum rather than
hold. The excess is rate-limited and the resolver's 8s per-attempt
deadlines expire, failing ResolveIPAddresses tests this branch never
touched. script/test now passes -p 1 so each gate is authoritative
while its package runs. Serialising is also net faster here, because
the retries it removes cost more than the lost parallelism: resolver
16-22s (was 30-36s under contention), watcher 12-13s (was 27-37s),
whole suite 39-46s against the 60s cap and the 90s -timeout backstop.

TestQueryNameserverIP_UnreachableServer is dropped rather than fixed.
It asserted that a query to an RFC 5737 documentation address comes
back non-OK with no records, which does not hold in the build
environment: that network transparently intercepts all UDP/53 traffic
regardless of destination and answers it locally, so the query returns
StatusOK with 9 real records for example.com. Verified directly with
dig @192.0.2.1 inside the build network. The mock DNSClient that used
to force this classification is what this PR removes, and a live
substitute would only be testing the sandbox's network behaviour, so
the coverage gap is recorded in a comment where the test was.

Verified: three consecutive `docker build .` runs green, after three
consecutive failures without these changes.

TESTING.md

@@ -31,6 +31,13 @@ assertions and sensible timeouts, not from mocks.
 - Watcher change-detection tests seed a synthetic *previous state*
   and compare it against fresh live lookups; the DNS side is never
   faked
+- Live query concurrency is bounded per package (`liveGate` in
+  `internal/resolver`, `liveWatcherGate` in `internal/watcher`) so
+  parallel tests do not burst at the root servers
+- Those gates are package-scoped and therefore per test binary, so
+  `script/test` also passes `-p 1`: with both live-DNS packages
+  running at once the gates sum instead of holding, and the resolver's
+  per-attempt deadlines start expiring
 
 ### What NOT to do
 
@@ -41,4 +48,8 @@ assertions and sensible timeouts, not from mocks.
 - **Do not remove `-count=1` from `script/test`** — Go's test cache
   replays a previous run's output without querying anything, so a
   cached pass is not evidence that live resolution works
+- **Do not remove `-p 1` from `script/test`** — running the live-DNS
+  packages in parallel oversubscribes live DNS past what their gates
+  bound, which surfaces as unrelated resolver tests failing on
+  expired deadlines
 - **Do not modify linter configuration** to suppress findings

internal/resolver/resolver_test.go

@@ -8,7 +8,6 @@ import (
 	"sort"
 	"strings"
 	"testing"
-	"time"
 
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
@@ -518,41 +517,15 @@ func TestQueryAllNameservers_ContextCanceled(t *testing.T) {
 	assert.Error(t, err)
 }
 
-// ----------------------------------------------------------------
+// The resolver's transport-failure classification (StatusTimeout /
-// Unreachable nameserver tests
+// StatusError for a nameserver that does not answer) is deliberately
-// ----------------------------------------------------------------
+// not covered here. Forcing it required the mock DNSClient this
-
+// change removes, and a live substitute is not available: the build
-func TestQueryNameserverIP_UnreachableServer(t *testing.T) {
+// environment transparently intercepts all UDP/53 traffic and answers
-	t.Parallel()
+// it locally, so a query to a black-holed address such as an RFC 5737
-
+// documentation address comes back StatusOK with real records. See
-	r := newTestResolver(t)
+// TESTING.md; restoring this coverage needs a mechanism that is
-
+// neither a mock nor dependent on the sandbox's network behaviour.
-	ctx, cancel := context.WithTimeout(
-		context.Background(), 10*time.Second,
-	)
-	t.Cleanup(cancel)
-
-	// 192.0.2.1 is an RFC 5737 documentation address: no
-	// nameserver can exist there. Depending on the network
-	// path the queries either time out (silent drop) or fail
-	// fast (ICMP unreachable), so accept any non-OK status;
-	// the resolver must return a classified response with no
-	// records rather than an error or a hang.
-	resp, err := r.QueryNameserverIP(
-		ctx, "unreachable.test.", "192.0.2.1",
-		"example.com",
-	)
-	require.NoError(t, err)
-
-	assert.NotEqual(t, resolver.StatusOK, resp.Status)
-
-	totalRecords := 0
-	for _, values := range resp.Records {
-		totalRecords += len(values)
-	}
-
-	assert.Zero(t, totalRecords)
-}
 
 func TestResolveIPAddresses_ContextCanceled(t *testing.T) {
 	t.Parallel()

internal/watcher/watcher_test.go

@@ -191,12 +191,40 @@ type testDeps struct {
 	config      *config.Config
 }
 
+// liveWatcherConcurrency caps how many of this package's tests may
+// be driving live DNS at once. Every test here runs a full iterative
+// resolution, and the package is entirely parallel, so without a
+// bound all of them burst against the same root servers in the same
+// instant and get rate-limited — the fan-out pathology that
+// internal/resolver/livedns_test.go exists to prevent. That gate is
+// package-scoped and so cannot reach across into this package's test
+// binary; this is its counterpart here.
+const liveWatcherConcurrency = 3
+
+// liveWatcherGate bounds concurrent live-DNS watcher tests. It has to
+// be package scoped: the point is that every parallel test shares it.
+//
+//nolint:gochecknoglobals // package-wide live query rate limit
+var liveWatcherGate = make(chan struct{}, liveWatcherConcurrency)
+
+// acquireLiveSlot blocks until this test may run live DNS, releasing
+// the slot when the test ends.
+func acquireLiveSlot(t *testing.T) {
+	t.Helper()
+
+	liveWatcherGate <- struct{}{}
+
+	t.Cleanup(func() { <-liveWatcherGate })
+}
+
 func newTestWatcher(
 	t *testing.T,
 	cfg *config.Config,
 ) (*watcher.Watcher, *testDeps) {
 	t.Helper()
 
+	acquireLiveSlot(t)
+
 	deps := &testDeps{
 		portChecker: &mockPortChecker{},
 		tlsChecker:  &mockTLSChecker{},

script/test

@@ -19,15 +19,26 @@
 #
 # -timeout 90s is a deliberate backstop above the 60s hard cap on
 # suite duration. Do not lower it.
+#
+# -p 1 runs one test package at a time, and is load-bearing. Live DNS
+# is a resource outside the process: the concurrency gates that keep
+# this suite from bursting at the root servers
+# (internal/resolver/livedns_test.go, internal/watcher/watcher_test.go)
+# are package-scoped, so each one only bounds its own test binary. Go
+# runs package binaries in parallel by default, so with both live-DNS
+# packages in flight at once their gates sum instead of holding, the
+# root and TLD servers rate-limit the excess, and the resolver
+# package's per-attempt deadlines expire. Serialising packages is what
+# makes each gate authoritative while its package runs.
 set -eu
 
 ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
 
 main() {
     cd "$ROOT"
-    go test -count=1 -race -timeout 90s -cover ./... || {
+    go test -count=1 -p 1 -race -timeout 90s -cover ./... || {
         echo "--- Rerunning with -v for details ---" >&2
-        go test -count=1 -race -timeout 90s -v ./... || true
+        go test -count=1 -p 1 -race -timeout 90s -v ./... || true
         exit 1
     }
 }