GitHub Actions outage left some CI events unrecoverable

GitHub Actions ran degraded for 10.5 hours starting August 6 at 15:22 UTC, per GitHub Status. Job completion dropped to 30-40% before recovering to 99% by 02:04 UTC on August 7, and the outage hit both hosted and self-hosted runners.

Anyone running review bots, deploy gates, or autonomous agents off those triggers inherited a gap they can't audit from GitHub's side.

Copilot code review and the Copilot coding agent failed or delayed alongside the core outage. Both depend on the same runners that were degraded.

Some push and pull-request events were never processed, and GitHub says they can't be replayed. Workflows tied to those events simply didn't run. No error, no retry, no second attempt.

The incident report doesn't say how many events were lost, or whether any of Copilot's coding-agent runs are among the ones gone for good.

Automatic recovery mechanisms are coming in future releases, GitHub says, but that's a roadmap item, not something that exists now.