There are two ways (in reality they are the same though( to report on things that are no longer there
1. Incessant polling - you need to ask your gateway cluster members "Are you still there? Are you healthy? Do you have enough disk space/memory/compute cores?" And - the most crucial of all questions: "Can you still see the proxy?" Without a working proxy your gateway is the digital equivalent of a door stopper.
2. Listening to the heartbeat of the cluster members. Have them report their telemetry every five minutes and stuff the data into a streaming dataset. But then you still need to understand when these heartbeats stop which takes you back to point 1.

This is all only looking at the physical side. On the logical side you can have a whole host of other issues like expired credentials, or changes in the data source.
In conclusion - if you want to be notified of gateway failures you have to work for it.