When executing `createOrUpdate()` on a node that is cached but has been removed from the database (by a different instance), the operation will fail, and continue to do so until the cache entry gets evicted or the cache gets manually cleared.
The exception message somewhat misleadingly says "update of ... failed, race condition?"
The problem has been fixed already in
OAK-7725, but that change is rather big and non-trivial to backport.
As a simpler fix for the maintenance branches we should invalidate the cache when the update fails, so that the system at least recovers and a retry will work.