docs(evidence): re-verify CI queue re-starvation item — already resolved (queue empty at 03:45Z)

.lane/reports/ci-starvation-rereverify-lane1-20260815.md

@@ -0,0 +1,54 @@
+# Lane report: CI queue re-starvation re-verification (lane 1, 2026-08-15, second pass)
+
+Branch: `ci/starvation-rereverify-lane1-20260815`
+Item: "The CI queue has re-starved past the closed 1126 acceptance: 9 runs queued up to 2h03m (18:00"
+
+## Verdict
+
+Already resolved — no repo change required. This is a re-confirmation of the
+diagnosis recorded in PR #214 (`ci/starvation-requeue-cleanup-lane1-20260815`)
+and re-verified in PR #217 (`ci/starvation-reverify-lane1-20260815`): the
+"9 runs queued up to 2h03m" observation was caused by PR #206's per-instance
+matrix routing its `tinystudio-io-verify1` legs to a label that no registered
+runner carries. PR #206 is closed (2026-08-14T22:36:18Z), its zombie legs are
+gone, and the queue is draining normally across both runners.
+
+## Live verification (2026-08-15 03:45Z–04:17Z, via GitHub API)
+
+- Runners registered and online:
+  - `netcup-rs2000-tinystudio-verify1` — labels `self-hosted Linux X64 vps-verify tinystudio-io`
+  - `netcup-rs2000-tinystudio-verify2` — labels `self-hosted Linux X64 vps-verify tinystudio-io-verify2`
+- Both workflows on main route through the shared `vps-verify` label:
+  - `.github/workflows/ci.yml` → `runs-on: [self-hosted, linux, x64, vps-verify]`
+  - `.github/workflows/secret-scan.yml` → `runs-on: [self-hosted, linux, x64, vps-verify]`
+- Queue state at first inspection (~03:45Z): **zero queued, zero in-progress**.
+- A concurrent 04:01Z wave (8 runs across PRs #91/#142/#194/#210/#211/#213/#218
+  and this branch) produced a transient backlog: 4 queued at peak, oldest wait
+  ~14m (04:01:12Z–04:15:28Z). It fully drained — zero queued by 04:15:28Z —
+  and PR #224's own `verify` (run 31863339003, on
+  `netcup-rs2000-tinystudio-verify1`) and `Gitleaks` (run 31863339025, picked
+  up at 04:10Z) both PASSED. Max queue wait observed: ~14 minutes, orders of
+  magnitude below the item's 2h03m.
+- Recent drain times:
+  - main pushes 03:43:00Z → done 03:43:16Z / 03:44:35Z (~16s–95s)
+  - PR branches 03:39:27Z → done 03:39:48Z / 03:41:07Z (~21s–100s)
+  - 03:01Z batch of 8 PR runs → all done 03:04:10Z–03:10:12Z (≤ ~9m, two runners engaged)
+- No open PR modifies `.github/workflows/*` (scanned all 14 open PRs by diff
+  name-only). No workflow runs exist on PR #206's branch.
+- PR #206: CLOSED, not merged.
+
+## Conclusion
+
+The item's premise ("the CI queue has re-starved") no longer holds as of
+2026-08-15 04:17Z: the queue fully drains (max observed wait ~14m during a
+concurrent 8-run wave, vs the item's 2h03m), both runners are online and
+sharing load through the shared `vps-verify` label, and the unregistered-label
+source (PR #206) is closed. PR #224's own CI completed green end-to-end during
+the observation window. Item closes as already resolved / no-op, consistent
+with PR #214 and PR #217.
+
+## Not done
+
+- No change to `.github/workflows/*` — main's shared-label routing is already
+  the correct fix for the original 1126 choke point.
+- No runner provisioning changes — runner labels are VPS-side out-of-band state.