Great feedback re the risks of physical metaphors! I'd agree a bad one does more harm than good. Lakoff points out that our linguistic conceptions typically compose-- so you can talk about both IDEAS as BUILDINGS and as CONTAINERS because the former is a special case of the latter-- and we lose that quickly when mapping to programming. I'm not sure if that's inherent in how abstract these things are or if it just hasn't been attempted on a deep enough level. (Do other control structures fit into the railway paradigm? Is it generative?)

Enjoyed your post, lots to unpack in the 'cleanliness' analogy, will think on it.