GitHub opens a critical Actions incident at 15:11 UTC, and a database primary failover does not fully mitigate

Sources: incident, incident API, discussion
Summary: GitHub opened incident y1t7p9fzrlj2, titled Incident with Actions, at 15:11 UTC on 2026-08-26 with impact critical, Actions at major outage and Pages at degraded performance at open. The published updates run 15:12 Pages degraded, 15:23 an issue identified with a database primary and a failover to a replica underway, 15:48 the failover briefly improved performance but did not fully mitigate, inbound traffic throttled and upstream Vitess issues under investigation, 16:14 believed identified and addressed with traffic ramping back slowly and some customers still seeing delays, 16:49 Pages operating normally, and 16:50 recovery being observed with delayed queues burning down and full completion expected within the hour. The incident carried no resolved entry at the last read, so it was still open.
Comments: Hacker News item 49450722 links incident hcbtzksccj2f, a separate and lesser event titled Disruption with some GitHub services at impact minor, open 15:09 to 16:07 UTC and resolved. The thread is discussion of the visible symptoms rather than of the Actions incident tracked here.
Why it matters: Actions is the CI path for a large share of public and private repositories, so an open critical incident affects builds and Pages deploys rather than a background service.
Follow-up: Track the resolution and the root cause analysis GitHub states it will publish.