A reliable machine can make its operator less practiced at the moment it finally fails.

The more reliable a machine, the less its operator practises. Skills fade through disuse, and the day the system finally fails is the day it hands control to someone who has not done this manually in years. Reliability quietly removes the competence that reliability was supposed to be backed up by.

This applies to anything you have automated: navigation, calculation, memory, decisions handed to a tool. Occasionally do it the hard way on purpose, when the stakes are low. It is inefficient by design, and it keeps a capability alive that will only ever be needed at the exact moment you cannot acquire it quickly.