This is what OCaml does: you can annotate function calls with [@tailcall] to result in compiler diagnostics if the call in fact isn't in tail call position.
tailrec is more restricted: it only supports simple tail recursion (i.e. calling the same function in tail position) that can trivially be rewritten into a loop. Scala isn’t really in a position to optimize mutual tail calls, since those don’t in general collapse to loops.
