Confirmed impact
GitHub’s incident findings place the Actions disruption between August 6, 15:05 UTC and August 7, 00:14 UTC: 9 hours, 9 minutes. Hosted and self-hosted runners were affected. At peak, 71% of workflow runs suffered infrastructure failures; 75% of the remaining runs were delayed over five minutes. These figures describe workflow runs, not all GitHub users.
Cause and recovery
A deployment exposed capacity and concurrency weaknesses. Expanding capacity and limiting incoming work helped services recover, but a separate assignment bug left runners retrying invalid jobs. Further fixes cleared those queues. Some ARC runners required manual recovery after the wider incident ended.
Key times
| UTC | What the time represents |
|---|---|
| Aug 6, 15:05 | Impact began, according to the later findings. |
| Aug 6, 15:22 | First public investigation update. |
| Aug 7, 00:14 | End of the confirmed Actions impact window. |
| Aug 7, 02:04 | Incident marked resolved on the status page. |
The resolution-post time is later than the impact end. Neither establishes an individual customer’s downtime.
If a workflow remained stuck
GitHub’s incident support thread contains recovery instructions for affected ARC runners and workflows that could not be canceled. Check the applicable instructions and your workflow’s state before retrying a deployment; avoid duplicating a job that already completed.
For a problem happening now, check GitHub’s current status and the Actions troubleshooting guide.
Correction
This reviewed version replaces an automated archive that reported 10 hours, 20 minutes and mixed updates from separate incidents. The timeline above is tied to one official Actions incident. The existing report URL is preserved.