This incident was because configs for setting up the runner user were not applied during a routine runner binary upgrade. This issue affected ~5% of our users.
The resolution required a rebuild and rollout of the affected images.
This incident has been resolved.
