[Mikola Lysenko (mikolalysenko)]

Fixes #1795

Problem

release-readiness is a ~10 s job (checkout, two ci-reuse-*.py probes, release-lint.sh) on depot-ubuntu-24.04-4, and coverage, e2e-build (and through it the e2e fan-out), test, node-addon, docker-base and the cargo/yarn legs all needs: it. In Depot slow-start episodes (#1512) its boot sits at the head of the critical path. From the issue (16:18–20:16 UTC, 20 runs): boot p50 31 s, p90 126 s, max 307 s; merge_group run for #1748 reached ci-ok at 11.48 min, with release-readiness alone finishing at 5.32 min.

Change

release-readiness runs on ${{ vars.RELEASE_READINESS_RUNNER || 'ubuntu-latest' }}, mirroring what #1747 did for ci-ok. The job only needs git, python3, bash and gh — all on the hosted image. Setting the repo variable RELEASE_READINESS_RUNNER (e.g. depot-ubuntu-24.04-4) moves it back with no merge. No needs: edge, output or step changes. scripts/tests/test_ci_scheduling.py gets a test pinning the runner, alongside the existing ci-ok one.

Expected saving

From the issue's replay: merge_group #1748 11.48 → ~8.7 min; across 19 runs p90 −1.1 min, max −2.8 min; nothing in fast-host periods. ~0.1k Depot Linux job-min/day moves to free hosted minutes. No macOS/Windows change.

Measured result

This PR's run 40764064008 (pull_request, all 51 checks green):

	baseline (issue, 20 runs on Depot)	this PR (`ubuntu-latest`)
`release-readiness` queue (created → started)	—	2 s
boot (started → first step)	p50 31 s, p90 126 s, max 307 s	1 s
work	5–23 s	7 s
finished at (min after run start)	p50 0.93, p90 2.80, max 5.32	0.17

`coverage` and `e2e-build` started right after it (e2e-build done at 1.40 min, coverage (2) at 4.83); `ci-ok` at 5.02 min. Total 44.7 job-min over 44 jobs. Depot was fast during this run (clippy boot 6 s), so this run shows the gate's floor rather than a slow-start saving; the saving shows in slow-start periods and in the merge queue, which the profiler will verify after merge.

Where each test runs

Unchanged. No job, step or test moved or skipped; only the host of one gate job changed. Required checks ci-ok and clippy keep their names and ci-ok still needs every job (test_required_verdict_includes_every_job).

Validation

• python3 -B -m unittest discover -s scripts/tests: 376 tests OK.
• actionlint 1.7.7 and zizmor --offline on ci.yml: same findings as origin/main (37 / 59, all pre-existing), none new.

Risk

If the hosted Ubuntu queue backs up as in #1372, this gate waits too; since #1747 the hosted queue for ci-ok has stayed at 1–7 s. Escape hatch: the RELEASE_READINESS_RUNNER variable. The previous DISABLE_DEPOT_RUNNERS switch is moot for this job (it already selected ubuntu-latest).

https://claude.ai/code/session_011f9aeRFXYtAsVJCfxwhyiz

---

Generated by Claude Code

[Mikola Lysenko (mikolalysenko)]

bugbot run

---

Generated by Claude Code

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 0e49c68. Configure here.}

[Mikola Lysenko (mikolalysenko)]

Ready for review at 0e49c686 (burn-down agent).

• CI: 51/51 (32 success, 19 skipped) check runs green or skipped on 0e49c686; no main-wide failures.
• Bugbot: reviewed 0e49c686, no findings; no open review threads.
• Mergeable, no CHANGELOG.md change.
• Slack announcement not sent this run (Slack send unavailable); the next run retries.

---

Generated by Claude Code