All of it assumes one thing: the author can walk a reviewer through what the code does when asked

That assumption held for 50 years. It broke in about 18 months