🔴 CI Failure: PR Fast CI — 87296a7

[github-actions]

PR Fast CI failed on gh-readonly-queue/main/pr-560-149fc12dcf12f1fe45cbf380478b3d673f184c59

Field	Value
Workflow	PR Fast CI
Run	#1197
Attempt	2
Branch	gh-readonly-queue/main/pr-560-149fc12dcf12f1fe45cbf380478b3d673f184c59
Commit	87296a7
Triggered by	@githubrobbi

Filed by ci-failure-notify.yml after the 60-second auto-rerun-transient window: auto-rerun-transient dispatched a retry but attempt 2 also failed → persistent flake or real regression.

[github-actions]

PR Fast CI failed on gh-readonly-queue/main/pr-563-8b1c74b3911efa8fcf6b58d8439f8127b6b328b2

Field	Value
Workflow	PR Fast CI
Run	#1204
Attempt	2
Branch	gh-readonly-queue/main/pr-563-8b1c74b3911efa8fcf6b58d8439f8127b6b328b2
Commit	e14488e
Triggered by	@githubrobbi

Filed by ci-failure-notify.yml after the 60-second auto-rerun-transient window: auto-rerun-transient dispatched a retry but attempt 2 also failed → persistent flake or real regression.

[github-actions]

PR Fast CI failed on gh-readonly-queue/main/pr-565-fad111d11b315fdf057e734ccf61527b22acaf98

Field	Value
Workflow	PR Fast CI
Run	#1213
Attempt	2
Branch	gh-readonly-queue/main/pr-565-fad111d11b315fdf057e734ccf61527b22acaf98
Commit	8f04be8
Triggered by	@githubrobbi

Filed by ci-failure-notify.yml after the 60-second auto-rerun-transient window: auto-rerun-transient dispatched a retry but attempt 2 also failed → persistent flake or real regression.

[githubrobbi]

Closing as stale — the cause is fixed and the code has left this repo.

Root cause

All three occurrences (PRs 560, 563, 565) are the same single test:

uffs-content job::tests::concurrent_lease_runs_actually_overlap_and_never_interleave_a_candidates_frames
elapsed 483.4449ms should be well under the fully-sequential estimate 600ms
   -- lease runs (drives) do not appear to be running concurrently

A wall-clock-ratio assertion (elapsed < sequential_estimate * 3/4) against a 100ms-per-operation fixture. On a loaded or throttled GitHub-hosted Windows runner, thread::sleep(100ms) alone can exceed the budget, so the test failed on machine speed rather than on concurrency. Not a product regression.

Why it is gone

1. The assertion was replaced. fad111d11 (feat(content): VSS-snapshot content-ingest pipeline for Docenta (UFI.0-UFI.6) #563) introduced it; 5c50616f7 (chore: release v0.6.28 — ship pipeline auto-commit #564) replaced it with assert_any_two_intervals_overlap, which records each operation's (start, end) and asserts that two intervals genuinely overlap. Overlap is true or false independently of how fast the machine is, which is what "concurrent" actually means.
2. The crate left the repo. 97ff57375 (chore: remove uffs-content family — moved to uffs-products #567) moved the uffs-content family to the private uffs-products repo. There are no content crates under crates/ any more.

Why it kept firing after the fix landed

The queue branch was gh-readonly-queue/main/pr-565-fad111d11b31…. That suffix is the base commit — fad111d11, the commit that introduced the bad assertion, before 5c50616f7 fixed it. The merge queue tested the PR against a stale base, so it ran the old assertion even though main already carried the new one. Worth recognizing if a similar merge-queue-only failure shows up again: check the base commit in the queue branch name before assuming main is broken.

Note for bisects

1191e405f — "fix(test): stop asserting concurrency via a wall-clock threshold (#566)" — is an empty commit: its tree is byte-identical to its parent. The fix content had already arrived with the release auto-commit 5c50616f7 (#564), so the PR squashed to nothing. The commit that claims the fix is not the commit that made it.

Downstream

uffs-products inherited the fixed version: the tree at 97ff57375^ contains assert_any_two_intervals_overlap, not the wall-clock assert. Nothing followed the crate over.

Every PR Fast CI run since (PRs 568, 569, 571, and main) is green.