Yeah - like, why do they not have an "HA" cluster, where they patch and reboot a subset at a time... Something this critical should have some sort of failover/load-balancing in-play...