If you rely on an optimization like this, that's not a code smell as much as a code-sewer:

This is one of those optimizations that a very clever compiler might be able to do interprocedurally give a certain input for example.

I expect the compiler would optimize the tail recursion into a loop, so the likely worst case is not that you blow the stack but that you spin for a while.

Not really, any expectations here are unsupported by C standards. Compiler is, by spec, allowed to do anything here.