I had to disable TTMServerMessageUpdateJob on the site (they get pushed but not popped) since they where absurdly slow, average 3-4 hours. They took 88% of job running time for .02% of jobs. After disabling them and restarting the runners, the run rate increased by ~10x and the huge queue backlog is starting to drop.
Apparently, these normally take minutes...I'm not sure what the issue is. It would help if we had CLI profiling.
Version: unspecified
Severity: normal
See Also:
https://bugzilla.wikimedia.org/show_bug.cgi?id=48164