There are lots of reasons that your instances in an AZ or cloud could all go down, other than the AZ's/cloud's operator self-reporting failure! For instance, wh

There are lots of reasons that your instances in an AZ or cloud could all go down, other than the AZ's/cloud's operator self-reporting failure! For instance, what if someone wanted to do a rolling update starting in one AZ, and it starts to saturate a limited resource in that AZ halfway through, that doesn't affect the overall AZ but does affect your instances therein? Your internal health checks, rebalancing, and monitoring should be what you use to determine your health and trigger rebalancing.