Theoretically the REP case can be optimized if it is known not to fault, but that is one legitimate use case of this instruction. This is just to ensure that we have this instruction covered, it hasn't been encountered in a real program yet.