Very much this. It's like with data-recovery. If things are extra-broken you don't try to fix the problem from within; you bring external order (A rescue disk

Very much this. It's like with data-recovery. If things are extra-broken you don't try to fix the problem from within; you bring external order (A rescue disk, taking out the problem storage disk and mounting it to another system) and then fix the problem from a working host environment.