This is merely a question of fidelity. One can imagine a simulator which produces low-fidelity everything and only increases fidelity gradually in isolated regi

This is merely a question of fidelity. One can imagine a simulator which produces low-fidelity everything and only increases fidelity gradually in isolated regions as the human observers peer at those regions. And, there is nothing to prevent results being continuously retroactively computed (just in time) as new 'discoveries' are made.