Are you aware if they do the failover, then after it is successful perform maintenance? Or do they take the just stop the master server to do maintenance and let failover happen automatically?

The first method would eliminate downtime. The second method would have that up to 3 minute downtime.

Also, in my basic testing, I only had 1 out of 10 failovers be completed within 15 seconds. The rest took between 2 and 3 minutes. The longest one was slightly over 3 minutes.