I think the idea is that there are error recovery semantics that: 1. Determine the last sane state of the system, and work forward from there. (Read the servo

I think the idea is that there are error recovery semantics that: 1. Determine the last sane state of the system, and work forward from there. (Read the servo position and try to go from there) 2. Have a the "recovery" routine to reset the system. (Take all positions to "zero") 3. Just stop. (Yes, I know this can be bad). And ask a human for help.