Is there a way to basically do what prometheus.ex does normally (including custom instrumentation for the app) but make it go through Telemetry?
Everyone who has ever promoted Erlang to me insisted that such things aren’t required and that Erlang magically recovers any crashes and so none of this matter. Were they wrong?
“Let it crash” is about keeping unexpected errors from corrupting the state of the system. You let that one little process crash and restart in a good state. It doesn’t mean your code shouldn’t handle expected situations. Monitoring helps you expand your expectations based on experience.