TestServiceLifecycle_Stop/user_stop:_forceful (pkg/lifecycle) is flaky in CI. It has spuriously failed the test job on at least two unrelated PRs (#2516 Docker fix, #2520 SIGTERM fix), each passing on re-run. Locally it passed 10/10 with -count=10.
The subtest is the forceful-stop path, which is timing/goroutine-scheduling sensitive under CI load. It's a CI-reliability problem — it blocks unrelated PRs and trains maintainers to re-run red, which is how real failures get missed.
Action: make the forceful-stop assertion deterministic (synchronize on the actual stop signal rather than a timing window), or quarantine it with a tracked skip until fixed. In the same lifecycle-concurrency territory as #1659 and the SIGTERM force-stop follow-up (#2519).
TestServiceLifecycle_Stop/user_stop:_forceful(pkg/lifecycle) is flaky in CI. It has spuriously failed thetestjob on at least two unrelated PRs (#2516 Docker fix, #2520 SIGTERM fix), each passing on re-run. Locally it passed 10/10 with-count=10.The subtest is the forceful-stop path, which is timing/goroutine-scheduling sensitive under CI load. It's a CI-reliability problem — it blocks unrelated PRs and trains maintainers to re-run red, which is how real failures get missed.
Action: make the forceful-stop assertion deterministic (synchronize on the actual stop signal rather than a timing window), or quarantine it with a tracked skip until fixed. In the same lifecycle-concurrency territory as #1659 and the SIGTERM force-stop follow-up (#2519).