Repository navigation
tests: deflake pre-existing flaky Go and E2E tests (index-loading 503, relay-info race, start-up race, fixture age) #567
Description
Pre-existing flaky tests reported by reviewers on 2026-10-24. All of them reproduce on master and none is caused by the PR under review.
| Test | Symptom | Rate |
|---|---|---|
TestHandleNodePaths_SortByRecency_1145, TestHandleNodePaths_PrefixCollision_1352_FallbackBranch |
503 "index loading" under CPU load | 1/30 on master in isolation |
TestRepeaterRelayInfoMap_ConcurrentIngestAndEviction |
fails | 2/150 on master |
TestServerProcess_NormalStartDoesNotWait |
start-up/SIGTERM race (CI Go Server Tests 3/3 on #555) | seen once in CI |
test-e2e-playwright.js "Version info lives on Perf dashboard" |
fails locally when the server cannot resolve a git commit from its cwd; passes in CI | environment-dependent |
| observer-IATA mobile E2E | fails once the freshened fixture is about 30+ min old | time-dependent |
test-issue-531-region-url-e2e.js section 2 |
API-count timing (tracked in #556) | 1/10 under load |
Goal
Each test should be deterministic: wait for readiness (index loaded) instead of racing it, avoid wall-clock fixture ageing, and remove cwd/git dependence. Every fix needs evidence of N≥30 green runs under CPU load, and a test that still fails on the real bug.
Activity
test(server): reproduce the acct113 fixture writer failing on a serve…
test(server): open the acct113 fixture like the ingestor opens its DB (…
test(e2e): measure the RF Health tile in one task (#567)
dborup commentedon Oct 24, 2026
Fixed by #571 (merged as
b80d453e87b9cdd73dfbf9116d6808888b20add6).