Distillation

Date: Sep 5, 2026

• 🟡 (18:59) .github/workflows/ci-poller.yml lines 404–415 include a Sync poller variable with pending issue state step running always() after successful poller-token; it uses GH_TOKEN: ${{ steps.poller-token.outputs.token }} and sets repository variable CI_POLLER_HAS_PENDING to "false" when steps.remaining.outputs.count == "0" or "true" otherwise via gh variable set CI_POLLER_HAS_PENDING -R "$GITHUB_REPOSITORY" -b ....
• 🟡 (18:59) .github/workflows/ci-poller.yml lines 417–440 self-dispatch ci-poller.yml for rapid CI rechecks because GitHub */5 cron may drift to 30–40 minutes under load; expected GitHub Actions startup latency gives approximately 30–60 seconds between checks, while concurrency permits 1 running and 1 queued workflow.
• 🟡 (18:59) The ci-poller self-dispatch chain is capped at 60 attempts (approximately 30 minutes); at attempt >= 60, it emits ::warning::Max self-dispatch attempts (60) reached. Relying on cron fallback. and exits 0. Cron continues checking afterward, and a subsequent accepted-label re-add or workflow_dispatch restarts the chain.
• 🟡 (18:59) The Self-dispatch if issues remain pending step runs only when always(), steps.token.outcome == 'success', steps.remaining.outcome == 'success', and steps.remaining.outputs.count != '0'; it uses GH_TOKEN: ${{ steps.token.outputs.token }}, derives ATTEMPT from ${{ github.event.inputs.attempt || '0' }}, increments it, and invokes gh workflow run ci-poller.yml -R "$GITHUB_REPOSITORY" -f attempt="${attempt}".