I was objecting to your assertion that a "deadlock" would necessarily cause the autonomous nervous system to fail.

Since computers don't do that, why do you think that evolved systems - which have billions of years to build up robustness - must do that?

Your objection to the "homunculus problem" is not relevant. A watchdog does not need to be a "brain" in any but the most poetic sense. It can be switch on a timer, where the timer is reset by the system.

Computer can have deadlocks on many levels. I could write a multi-threaded program which deadlocks. Even though that's frozen, the rest of the computer could act like normal.

Since you didn't specify which sort of deadlock you were looking for, I pointed to ones which occur at the most recently evolved parts of the brain.