Actually, that's an interesting question, Integral.
I had been implicitly presuming the same thing; I was talking about how "easy" it is to reemit the energy in a particular direction, but kept tacit the law of large numbers which (kinda sorta) states that the net result is predominately the easiest one.
But... some of the interesting QM experiments seem to say otherwise. For instance, in the famous experiments involving half-silvered mirrors, photons are said to either be transmitted through the mirror, or to be perfectly reflected via incidence = reflection...
I guess (I think) it falls down to the fact that the photon is really the center of probability for a wave packet, so by definition it follows perfectly the expected behavior.
Hurkyl