- Turn the new
health_snapshot_timingevidence into a concreteDevServer.healthfix.
- Live trace from the restarted shell showed:
health_snapshot_enterimmediatelyhealth_snapshot_timingwith:dirty_ms = 138162.33hash_ms = 0.289lock_wait_ms = 0.001
- That isolated the remaining delay to the shell dirty-state probe.
- Replaced the unbounded shell dirty check with a bounded helper:
- old path:
git status --porcelain --untracked-files=all - new path:
git diff --quietandgit diff --cached --quiet - both guarded by a timeout
- old path:
- Added explicit
shell_source_dirty_checkreporting so live health responses can distinguish:oktimeout- git exit-code failures
- Targeted tests: passed
- Full suite:
173 passed
- Restart the hot shell and re-run live
DevServer(mode="health")to verify that the timeout is gone or at least degrades into a fast, diagnosable timeout-status response.