> In that case, the database will perform a failover in those instances so there isn't downtime.

Have you done that? I have and there's been downtime every time (~5m).