Fully agree here. Catastrophic failures can slip in silently on both ends of the curve -- if Joe doesn't get an email OR Joe gets too many emails, it's still bad. In my experience, beefing up logging to be able to do distributed tracing is an absolute must to prevent debugging pain.