Memory leak on EventStoreTask makes CPU soar and hangs Xwiki until restart

See https://forum.xwiki.org/t/cpu-100-for-a-while-xwiki-cant-seem-to-recover-afterwards-not-sure-how-to-debug/11574/18 for the history of this ticket.

We're running Xwiki in a docker swarm, we're running mysql-tomcat variant.

We've upgraded from Xwiki 13.9.x (not sure anymore) to 14.10 because of the text4shell exploit.

Right away we had this issue where the CPU spikes and leaves us with a non-working Xwiki (going to a page stays on 'pending'). Where a restart fixes this. This isn't tied to time or anything, sometimes it happens every hour sometimes Xwiki works for days.

After a while I opted to start a clean Xwiki and start exporting and importing all our pages, this did not fix the issue.

When doing a heap dump we can see a large amount of tasks from the EventStoreTask (642.847 entries which is 87.89% of the 2gb dump!):

We are almost certain this is the problem comes from user interaction as we have a test xwiki running with the same configs which stays untouched which still works.