GitHub outage lasted 7 hours and 47 minutes

GitHub’s August 17 outage lasted 7 hours and 47 minutes, after a critical component in its Central US data center failed to scale at a traffic peak and a Copilot client retry loop increased load during recovery. In GitHub’s postmortem, the company says the incident disrupted authentication, Actions, APIs, pull requests, issues, and Copilot; its immediate fixes include consistent retry limits, retry budgets, and variable timeouts. Teams building agentic delivery systems should treat client retries as part of the reliability boundary, alongside GitHub’s pull-request control-plane design, not as a harmless SDK detail.