[ci-maintainer] Flaky test: TestDeadSessionRestartsOncePerBackoffNotPerTick (pkg/watchdog) fails on scheduled Coverage Hourly runs

[kubestellar-hive]

CI Issue

TestDeadSessionRestartsOncePerBackoffNotPerTick in pkg/watchdog (reconciler_test.go:1113) failed a scheduled Coverage Hourly run on v4 with:

--- FAIL: TestDeadSessionRestartsOncePerBackoffNotPerTick (5.00s)
    reconciler_test.go:1113: want 3 restarts across 30 sweeps (the backoff ladder)
FAIL	github.com/kubestellar/hive/pkg/watchdog	5.011s

The 5.00s duration suggests a wall-clock/timing dependency: under coverage-instrumented, loaded CI runners the sweep loop doesn't complete the expected 3 backoff-ladder restarts within the test window.

Evidence

• Workflow: Coverage Hourly, run 33011673035 (2026-08-26T20:41Z, branch v4, schedule)
• Subsequent 7 scheduled runs passed — intermittent, load-sensitive
• Test landed recently with the self-healing watchdog work (RFC RFC: agent self-healing watchdog — k8s-style liveness/readiness reconciliation for hive agents #4665)

Recommendation

Gate the assertion on observed sweep completions rather than wall-clock (inject a fake clock or count sweeps deterministically), or widen the window so the backoff ladder can complete under coverage instrumentation. This test cluster (pkg/watchdog reconciler) is disjoint from the open hold-gated PRs (#4871 pkg/agent bob-kick, #4864 pkg/dashboard nil-logger, #4863 readAuditLogFile, #4032 proxy auth).

---

Filed by ci-maintainer agent (ACMM L4/L5 — hold-gated mode)

🐝 Hive Agent: ci-maintainer | Instance: hosted-available-oke-11-placeholder-r05x | SHA: unknown

— hive: agent=ci-maintainer backend=copilot model=claude-fable-5