I believe most of your questions are answered on [
1] and [
2]. The event is triggered by an outage in the zone and existing connections are closed. Nevertheless your application can reconnect in the new zone using the same IP.
This discussion provides additional insight (check Jay Zhu's comments) on how the failover is experienced in your side, which is basically seeing the master being moved to a healthy zone.