8.11 (motir-core) planningTargetLockGate real-concurrency race times out under loaded Vitest shard
Opened by Zhu Yue ·
Type: test · Executor: coding_agent · Repo: motir-core
Parent: MOTIR-3875
Discovered in: MOTIR-4042, PR #2494, CI run 33492578435, commit dd22871ff.
Reproduction
Run the Vitest suite with the integration planning tests enabled. In tests/integration/planning/planningTargetLockGate.test.ts, the real-concurrency race test runs five rounds of overlapping sessions and timed out at 15 seconds on Vitest shard 7/12.
Evidence
The failed job ran for 23m29s and reported 1 timeout with 1744 passing tests:
FAIL tests/integration/planning/planningTargetLockGate.test.ts
> 1 · the race, with genuine concurrency
> gives exactly one of two overlapping sessions the epic, and tells the other who has it
Error: Test timed out in 15000ms.
All other Vitest shards and CI checks passed. MOTIR-4042 changes only the static OpenAPI route and its filesystem-boundary guard, so the timeout is unrelated to the PR diff. This is related to completed card MOTIR-3736, but that card fixed the sibling tests/planning/planTargetLockService.test.ts; this integration test remains a separate uncovered flake.
Root cause / fix direction
The five-round database race is sensitive to loaded Vitest-shard scheduling and cleanup latency. Give this integration test an explicit, owned budget or isolate its setup/cleanup so legitimate concurrency assertions remain intact. Do not remove rounds, weaken assertions, or patch unrelated lock behavior.
Resolution
Open — reproduce under CI load, then make the test budget/fixture lifecycle deterministic.
Discussion
No comments yet.
Adding to this discussion signs you in on app.motir.co and brings you back to this request.