I like the observation. It is a corollary of Amdahl's law: if every step is slow there is no point optimizing any of them until you can optimize all of them.