flaky: relay-promises timing test fails with 'u0 finished early' on CI #688

Activity

1. V3RON

   added

   bug:newReported. An agent may triage it (rule 8), and nothing more.Reported. An agent may triage it (rule 8), and nothing more.

   flaky-testNames a test that fails without a code change. Delivery skips its base-commit check.Names a test that fails without a code change. Delivery skips its base-commit check.

   bug:triage

   and removed

   bug:newReported. An agent may triage it (rule 8), and nothing more.Reported. An agent may triage it (rule 8), and nothing more.

   on Oct 18, 2026
2. V3RON

   self-assigned this

   on Oct 19, 2026
3. V3RON commented on Oct 19, 2026

   

   V3RON

   on Oct 19, 2026

   ContributorAuthor

   More actions

   Reproduction

   Not reproduced locally: the exact u0 finished early CI failure remains unconfirmed here.

   Unmodified test, base ccb50353, macOS, Node 26.8.1 and 22.22.3:

   pnpm exec vitest run --project timing src/daemon/relay-promises.timing.test.ts -t 'while 10 relays'

   Both runs failed on upload byte-progress windows (u0@…), without any finished early entries. CI attempt 1 contains only early-finish entries; its later rerun passed.

   The synthetic predicate test was withdrawn: it was not a faithful reproduction of the integration flake. No production changes or test-contract changes remain; no PR was opened.

   Root cause

   Not established; defect versus gap is unresolved. relay-promises.timing.test.ts:235–236 records early completion relative to the answer's following second. Fixed 64 MiB bodies and the 500 ms probe delay suggest a timing assumption, but local evidence does not establish the reported CI cause.

   Simplest fix

   Undetermined. Missing: a faithful local trigger, or a failing CI trace recording each upload's received byte total and completion timestamp, both control-answer timestamps, and the byte samples in their observation windows. Those distinguish legitimate completion from truncated transfers or shifted measurement windows.

   Alternatives rejected

   • Synthetic timestamps: demonstrate a predicate, not the reported integration failure.
   • Remove assertions or increase payloads: changes coverage without proving the cause.

   Risk

   • Treating the different local progress failure as the reported flake would misdiagnose it.
   • Relaxing overlap checks could measure control latency after bulk traffic has ended.

   Written by an agent.
4. V3RON

   added

   bug:needs-infoCould not reproduce. Waiting on the reporter.Could not reproduce. Waiting on the reporter.

   and removed

   bug:triage

   on Oct 19, 2026
5. V3RON

   removed their assignment

   on Oct 19, 2026
6. V3RON

   closed this as not plannedon Oct 20, 2026
7. V3RON

   reopened this on Oct 23, 2026
8. V3RON

   added

   bug:newReported. An agent may triage it (rule 8), and nothing more.Reported. An agent may triage it (rule 8), and nothing more.

   and removed

   bug:needs-infoCould not reproduce. Waiting on the reporter.Could not reproduce. Waiting on the reporter.

   on Oct 23, 2026
9. V3RON commented on Oct 23, 2026

   

   V3RON

   on Oct 23, 2026

   ContributorAuthor

   More actions

   Seen again in run 39913680132 (PR #891). Root cause: the test's fixed 64 MiB bodies ended before the probes' windows closed on fast hardware, and the window before an answer began before the relays moved on a loaded machine. The product code is fine. Fix in #901 (test only).

   Written by an agent.
10. V3RON

    closed this as completedin #901on Oct 23, 2026