docs(pr): do not enqueue into a busy runner pool

[CybotTM]

The arming gate lists three conditions, all about review state. A fourth decides whether the entry survives at all.

check_response_timeout_minutes (default 5) starts at enqueue and covers the wait for a runner, not just the run. With the pool saturated the required jobs sit in queued with no runner assigned, and the queue discards the entry having executed nothing. The PR looks fine throughout — CLEAN, approved, no failing check — which is what makes it hard to read.

Retrying amplifies the cause. Each attempt spawns a full set of runs on a gh-readonly-queue/* branch, and a dropped entry does not cancel them, so every retry leaves more runs holding slots and slows the next attempt.

Measured in netresearch/ofelia:

Four-attempt retry loop on a loaded pool	dropped every time, 18 unfinished runs piled up
Same required checks in PR context	concluded in 2.0 minutes — well inside the 5-minute window
Three later PRs, enqueued only at ≤2 unfinished runs	merged on the first attempt each

So the timeout was not too short for the work; the entry never got to do the work. This adds the gate, the one call that reads the timeout when an entry drops on an already-quiet pool (then it genuinely is the setting), and the cleanup for one's own orphaned queue runs — scoped to your own PR, since other PRs thrash the same way and their runs are not yours to cancel.

[github-actions]

Dependency Review

✅ No vulnerabilities or license issues or OpenSSF Scorecard issues found.

Scanned Files

None

[Copilot]

Pull request overview

Updates the Git workflow “Auto-Merge / Merge-Queue Arming Gate” documentation to address a failure mode where merge-queue entries are dropped due to runner-pool saturation (timeout starts at enqueue and includes waiting for a runner), and to provide operational guidance to avoid amplifying the problem with retries.

Changes:

• Adds a new arming-gate condition to only enqueue when the runner pool is quiet.
• Documents how check_response_timeout_minutes affects merge-queue drops and how to inspect the ruleset parameters.
• Adds CLI snippets for estimating runner-pool load and canceling orphaned merge-queue runs for the current PR.

Suppressed comments (1)

skills/git-workflow/references/pull-request-workflow.md:1263

• gh run list only returns the most recent runs up to --limit. With a busy repo, older queued/in_progress runs can be pushed outside the limit by newer completed runs, causing this count to under-report unfinished work and potentially enqueue into a still-busy pool.

gh run list --repo "$R" --limit 25 --json status \

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated no new comments.

Suppressed comments (2)

skills/git-workflow/references/pull-request-workflow.md:1264

• The "quiet runner pool" gate counts non-completed runs from only the most recent 25 workflow runs. Because gh run list is recency-ordered, older still-queued/in-progress runs can fall outside that window, producing an undercount and potentially a false "quiet" signal. Use a larger --limit (or otherwise ensure the query covers all active runs) so the gate is reliable on busy repos.

gh run list --repo "$R" --limit 25 --json status \
  --jq '[.[]|select(.status!="completed")]|length'

skills/git-workflow/references/pull-request-workflow.md:1283

• This snippet uses xargs -r, which is a GNU extension and will fail on macOS/BSD xargs. Prefer a portable loop so the "cancel orphaned runs" step works across common dev environments.

gh run list --repo "$R" --limit 40 --json databaseId,status,headBranch \
  --jq '.[]|select(.status!="completed")
        |select(.headBranch|startswith("gh-readonly-queue/main/pr-'"$PR"'-"))|.databaseId' \
  | xargs -r -n1 gh run cancel --repo "$R"

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[Copilot]

Pull request overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated no new comments.

Suppressed comments (2)

skills/git-workflow/references/pull-request-workflow.md:1264

• This gating command only inspects the most recent 25 runs, which can undercount unfinished runs in a backlogged repo and lead to a false “quiet pool” signal. It also relies on $R being set elsewhere, so it isn’t copy/paste-ready as written.

gh run list --repo "$R" --limit 25 --json status \
  --jq '[.[]|select(.status!="completed")]|length'

skills/git-workflow/references/pull-request-workflow.md:1284

• xargs -r is a GNU extension and fails on macOS/BSD. Using a while read loop keeps the snippet portable while still skipping the no-IDs case.

  | xargs -r -n1 gh run cancel --repo "$R"