Relationships
↔ sibling #4526#4519 control_plane_schedule_state_repository_test: 'list skips a record whose read times out' fails under load because the 20 ms read bound also covers the good record
Opened by stack72 · 11/8/2026· Shipped 11/8/2026
What happens
ControlPlaneScheduleStateRepository: list skips a record whose read times out and keeps the rest (src/infrastructure/persistence/control_plane_schedule_state_repository_test.ts) fails under machine load with:
error: Error: timed out after 20ms
at Timeout._onTimeout (src/infrastructure/persistence/datastore_format_marker_reader.ts:187:21)It failed a verify-build run-tests step while two other verification workflows ran on the same host (swamp-club#4516, commit 5e22283f2), and failed again locally under --repeats 20 on both that branch and origin/main while that build was running. Run alone (the same step rerun, and four local runs of --repeats 20 on an idle machine) it passes.
Likely cause
The test builds the repository with { timeoutMs: 20 }, and that bound applies to every record read, not only the hung one. The good record's read goes through the real store, so on a loaded machine it can take longer than 20 ms, list() gets a timeout on the record it should keep, and the assertion [good] fails.
Suggested direction
Remove the dependency on wall-clock speed for the good read rather than raising the bound: for example, time out only the hung key (a fake clock or a per-key timeout in the test double), or give the good read a store that resolves synchronously, so the 20 ms bound only ever governs the read meant to hang.
Shipped
Click a lifecycle step above to view its details.
Sign in to post a ripple.