I understand all that. You don't seem to have understood my comment, which was asking questions about the suitability/fitness of the "#lang" mechanism, rather than questioning the idea of polyglot code on top of Racket.
I guess it seems to me like it hasn't been established whether the "#lang" mechanism is a brilliant invention, or something cheesy. My initial reaction is that it seems cheesy.
I'm not sure I'm following, then. You think the basic idea of polyglot programming on top of Racket is sound; but that's what #lang does, and is the interesting bit of what we're talking about. The fact that it's specified with a line at the top of the file is a boring implementation detail - albeit the correct decision in my view, since the information about what language a file is written in clearly belongs with that file - exactly the same reason we have shebangs. But it seems like a minor thing to pick on to deem it "not Racket's USP".
I guess it seems to me like it hasn't been established whether the "#lang" mechanism is a brilliant invention, or something cheesy. My initial reaction is that it seems cheesy.