A runaway process in our cloud SDK generation pipeline was not being terminated and consumed CPU on the shared generation service, causing other remote generation jobs to time out before completing or publishing. We deployed a fix that detects and force-terminates stuck processes, restoring normal remote generation.
