When one connector instance has connection issues (broken connection) - resoruce is marked as down -> but this affects whole cluster. Multithreaded task will slow down, freeze because other threads (and probably even the one with broken connection) will attempt to put resoruce "up" -> completeResoure operation takes tens of seconds + 2 modifications. Whole system frozen.
Can we somehow improve this situation?
Attached operation result for specific delta (one of more) - whole operation took 40s.