The blind BACKUP node is easy to miss until a failover test, when you suddenly have no logs from the box that just took over. Worth adding to a checklist: after applying the fix, verify from a remote site that syslog, NTP and package updates all work from the BACKUP while it is still in backup state, not only after it becomes MASTER. That catches a broken route before a real outage does.