GitHub says availability work cut database load and expanded Azure traffic, but September incidents and a missed migration target show gaps remain

GitHub says August and September work reduced shared-database pressure, improved overload controls, added GitHub Actions headroom, and moved more workloads to Azure; by October 7, 120 of 160 production services and 13 of 51 primary MySQL databases were deployed in Central US. The company also reported disruptions including a September 13 shared-database incident affecting issue creation, signups, APIs, token creation, and Actions, plus a September 28 Copilot code review incident caused by a service change that omitted full review results for some customers.

For teams building on GitHub APIs, Actions, or Copilot workflows, the reported overload controls and partial PR failure containment suggest better degradation paths, but the September incidents show why clients still need timeouts, backoff, and retry limits. The trade-off is that migrations and cleanup jobs can create needed capacity, yet GitHub’s paused Central US ramp and incomplete database migration mean regional resilience remains a work in progress.