Update: GitHub's 7.5-hour outage is over — but no root cause yet
The 7.5-hour meltdown spiked Actions, API and SAML auth to ~20% errors; Copilot recovered last, and a new runner-permissions issue is under watch today.
Copy markdown
It's resolved — 21:15 UTC, ~7.5 hours end to end
GitHub marked the incident fully resolved at 21:15 UTC on Aug 17, about 7.5 hours after it began near 13:40 UTC. Web and API traffic peaked around 20% errors, while archive and raw-content downloads hit ~50%. Actions, Pull Requests, Issues, Pages and Webhooks all went red.
Copilot auth was the last thing to heal
Copilot flipped to “major outage” at 14:31 UTC and its authentication failures lingered after core services came back — so flaky Copilot or SSO logins into the evening were the tail of this incident, not a new one. Enterprise SAML/OIDC, SCIM and Team Sync were hit too.
Recovery came in waves, not one clean fix
Trackers logged services recovering and then failing again: Git Operations reportedly went back down 17:30–18:23 UTC and the API wobbled 18:48–19:01 UTC. If your CI passed, failed, then passed on Aug 17, that was the incident — not your pipeline.
Still no root cause, and a fresh runner incident today
GitHub has published no postmortem — it only said it fixed “the problematic component.” Meanwhile a new incident, “intermittent failures in runner group and runner-related permissions pages,” went to Monitoring at 07:40 UTC on Aug 18. Check Actions permissions before you blame your own config.