Resolved. The expired credential was rotated and the service restarted at 06:53 UTC on 30 August, and dashboard widgets began loading normally again immediately. We have confirmed no further authentication failures since that time. This incident is being recorded after the fact, so the timestamp on this update reflects publication rather than recovery.

Resolved. The matching web client version finished deploying at 11:06 UTC and the application has been loading normally since. Total impact window was approximately 16 minutes (10:50–11:06 UTC), limited to the web application; the mobile app and API were unaffected and no data was lost. This incident is being recorded after the fact. We are changing our deployment process so that client and server changes to the same API field can no longer ship out of order.

The issue has been resolved.