The availibality setting I mention is this one:
Disabling this will not make it more or less stable, only give more correct information on end-devices.
Hope it helps with replacing the coordinator.
I have read that a zigbee mesh can be triggered to do a re-route, if the coordinator is powered of for more than 12 minutes. The mains should go into a “panic” state and redo their routing. Might help. I do not have personal experience with this.