> if you feel that your well-understood regular failure case is viable, then all your error handling can fall-through to that case.

This is my favourite line, because it generalizes the underlying principle beyond the specific BEAM/OTP model in a way that carries over well to the more common sort of database-backed services that people tend to write.