Repository navigation
id(ci): F11.4 regression — fast-lane median PR feedback is 19m09s over the last ten runs (exit: under 15m) #1401
Description
Regression found by the #1078 validation audit (F11.4 measurement, slice 4) — 2026-10-27
F11.4's exit criterion (.opencode/plans/phase-11-foundations.md): "median PR feedback for the fast lane under 15 minutes, measured over ten runs." It was closed as done in #551. Measured fresh today, it no longer holds:
Last ten completed pull_request runs of the fast lane (workflow id, .github/workflows/id-check.yml), 2026-10-27, run wall time (created → updated):
| Run | Wall time |
|---|---|
| 40490775952 | 16m32s |
| 40480522636 | 15m06s |
| 40474146822 | 20m40s |
| 40468237269 | 16m07s |
| 40459695080 | 17m12s |
| 40458150431 | 39m34s |
| 40453523293 | 19m18s |
| 40452862315 | 18m59s |
| 40449139999 | 27m10s |
| 40448244193 | 24m48s |
Median: 19m09s. All ten runs exceed 15 minutes (the fastest by 6 seconds). Spot-check of run 40480522636's jobs shows the wall time is real work, not queue lag: matrix jobs started ~30s after run creation and the slowest (test-shard-2) ended at run end. Success-only runs are no better (15m06s / 19m18s / 20m40s).
Acceptance criteria
- Median wall time over ten consecutive completed PR runs of the
idworkflow is under 15 minutes again, with the run list recorded on this issue. - The slowest matrix checks at measurement time are named (crash-gate splits and test shards are the current long tail) and the fix targets them specifically (further splitting, caching, or moving a check to the nightly lane with the composition argument recorded).
Found by: queue task 1259 (validation audit of #1078), slice 4. Parent audit: #1078. Original F11.4 item: #551 (closed). Path-to-Phase-12 queue: #1235.
Activity
developing-today commentedon Oct 27, 2026
CODE-492
dezren39 commentedon Nov 2, 2026
Claimed as queue task 2572 (triage). Revalidating F11.4 fast-lane measurement against fresh runs, then routing or closing with evidence.
dezren39 commentedon Nov 2, 2026
Triage verdict: confirmed but re-scoped (queue congestion, not check duration)
1. The issue methodology is obsolete at HEAD. Commit
285e425b7(Drew policy, 2026-11-02 15:28 UTC — "CI never runs on pull requests") removedpull_requesttriggers fromid-checkand 7 other workflows, and removedmerge_groupfromnix-ci. The ten-runpull_requestsample this issue measures can never recur — that workflow no longer runs on PRs.2. The underlying regression is CONFIRMED, and it is runner queue wait, not check duration. Fresh measurement 2026-11-02 ~23:05 UTC:
opencode-config-guard, static checks): median ~36m created-to-completed over the last 15 completedpull_requestruns (successes 31–43m).basic: OpenCode metered-provider guard): executed in 6s (started 23:01:35Z, completed 23:01:41Z) inside a 31.0m run wall — ~30.9m was runner queue wait.idpull_requestruns from this morning: 370–600m walls, likewise queue-dominated.3. F11.4 needs redefinition (Drew decision, out of scope). The exit criterion ("median PR feedback for the fast lane under 15 minutes, measured over ten runs") no longer maps to any existing workflow.
Routing: follow-up queue task enqueued to prove the congestion breakdown on current main and land/propose a remediation (no PR-trigger restoration, no F11.4 redefinition, no held/Drew-owned artifacts).
28 remaining items
developing-today commentedon Nov 3, 2026
agent-note:v1
state: reviewed (coverage loop-back)
Coverage pass 2026-11-03, loop-back review: this issue was re-reviewed (drift / overlap / already-handled / stale / blockers / ordering) so every open issue in the four repos is reviewed and in the system. Verdict: still live; in the system as task 7240 (triage, queued, priority 50) in repo-health-code-pilot. Labels audited and synced. If it is already handled at current main, its triage closes it with evidence; if oversized, the triage files the split; blockers belong on this issue.
Found via: coverage loop-back 2026-11-03
dezren39 commentedon Nov 3, 2026
🤖 Agent note · picked up
dezren39 commentedon Nov 3, 2026
🤖 Agent note · triage findings — premise retired by CI policy; closing as superseded
.github/workflows/id-check.ymlat current main triggers ONLY onpushto main (pathspkgs/id/**+ workflow/scripts), a nightlyschedule, andworkflow_dispatch. There is nopull_requesttrigger — the PR "fast lane" this issue measures no longer exists as designed.pull_requestruns of workflowidwere 2026-11-02 (runs 41466878772, 41466379274, 41466076942, 41461633547, 41455364479) — every onecancelled, consistent with PR CI being switched off that day. Every run since ispush-event (main tier).idworkflow under 15 minutes" — is unmeasurable under current policy: there are no PR runs to measure, by deliberate policy, not by regression. The 19m09s median from 2026-10-27 described a lane that was retired a week later. Closing as superseded by the main-only CI policy.idruns are all sittingqueued(backlog since ~16:03Z) — that is a separate infra problem, already recorded by task 8850's triage of issue main-broken: developing-today/code id @38eba55 — basic: nix flake check (biome-check) #5123 and the lane handoff snapshot; it is not this issue's acceptance and does not block this close.