Sounds good in theory. Until the bug that causes failure is also present in the monitoring system, and as such doesn't fail over to the backup system. AKA exactly what happened here to Google.