He assumes that there are no more bugs once you put tests before coding, and he also assumes that there is zero cost to putting tests before coding. (Or at leas

He assumes that there are no more bugs once you put tests before coding, and he also assumes that there is zero cost to putting tests before coding. (Or at least, he doesn't mention these things.) With assumptions like that you can prove anything. Even without queuing theory.