On September 3, some background workers that collect analytics repeatedly restarted. Bluesky analytics was most affected, so some customers may have seen delayed or outdated insights. We did not identify an interruption to publishing.

The workers ran out of memory available to Node.js, even though their containers had additional memory available. We increased the affected workers’ memory limits and adjusted their Node.js settings. Restarts stopped, and analytics processing recovered.

We have not yet established what caused memory use to rise at that time. We are investigating that question, reviewing memory settings across other workers, and improving how we detect this type of failure earlier.