[wayfinder]

A good reliable system pretty much has error handling every step of the way, down to crash-first programming at the very end.

It’s more possible to recover from errors higher in the system but you can’t predict every problem so you need layers of error handling at increasingly lower levels with different recovery expectations.