I've compared, again and again, clients' captures done three different ways, all digitally, and they all match digitally. The only times I've seen bounces be different have been when the CPU was overloaded and automation moves were not quite performed at the same time.
Except one time recently. The null test revealed the guitar reverb and vocal reverb were different! Hey, no surprise, different on each pass due to the randomization algorithms in the reverb chamber.
Ho hum.... It's real easy to say you hear differences. It's a lot harder to admit that you don't hear them.
Here's another one, SADiE's automation display of fader levels seems to differ by 0.01 dB sometimes between automation record and automation playback, because they compute the fader levels to X number of bits. Does this mean that captures are bogus? Of course not. But it does mean that two passes will not null. Will they sound different because of those 0.01 dB differences? Could you tell?
BK