use a more useful and accurate heuristic to get ::input_streams() for the auditioner