GitHub had six availability incidents in February 2026. The worst hit on February 2, lasting nearly six hours: a telemetry loss caused security policies to be mistakenly applied to backend storage accounts in GitHub's underlying compute provider, blocking VM operations entirely. Actions hosted runners, Codespaces, Copilot, CodeQL, Dependabot, and GitHub Pages all went down across all regions. The fix was a policy rollback, but the damage was done.
February 9 produced two separate outages totaling 2 hours and 43 minutes of degraded service, both traced to a single configuration change in a user settings caching mechanism. The change triggered mass simultaneous cache rewrites that exhausted connections in the Git HTTPS proxy, taking down github.com, the API, Actions, Copilot, and core Git operations. The second incident that same day was caused by a second rewrite source the first mitigation missed. SSH operations were unaffected both times. A Codespaces incident on February 12 peaked at a 90% failure rate across Europe, Asia, and Australia due to an authorization claim change in a core networking dependency, compounded by alerts that fired at the wrong severity level.
The report is worth reading in full for the technical chain-of-failure details, particularly how a telemetry gap cascaded into a platform-wide VM outage and how a single cache config change broke Git HTTPS twice in one day. GitHub names specific mitigations already shipped, including write amplification fixes, Git proxy auto-recovery, and updated alerting thresholds, plus open items still in progress with their compute provider.
[READ ORIGINAL →]