I would assume it's also slightly more efficient that pd doesn't have to check the vector size when processing the s~ / r~ functions.