[account42]

With LTO the compiler should be free to fudge the calling convention for most calls even between translation units.

Here is GCC doing that optimization for a static noinline function: https://godbolt.org/z/cn6Wz9Kvn

Similarly, compilers can also clone functions if it makes sense to propagate constants from call sites into the function. Example: https://godbolt.org/z/59z6xT75n

I'm sure there is more room for improvement. A perfect compiler would always optimized programs as a whole and only regard function boundaries as hints at best. In practice, you have to keep complexity in check somehow.