The issue was caused a combination of bad upgrade and stale task data that starved our database and prevented new events to be processed. We have fixed the issue are seeing upgrades go through again
