Perfectly? Surely that is not reasonable?

Even imperfectly, the problem remains - and is answered by the "the energy of the cancellation source counterbalances the noise energy."