Looks roughly like how we do it as well with one exception - we don’t crash on “programmer errors” because we don’t maintain any local state in the node process

Looks roughly like how we do it as well with one exception - we don’t crash on “programmer errors” because we don’t maintain any local state in the node process. If all state lives in your DB, Redis, etc. it’s perfectly ok to have a catch all error handler that logs errors, sends it to Sentry, etc.