Thanks Karvala, this is a very interesting discussion.
I understand there is normally a lot of processing involved with most sampled pianos which have relatively few actual samples. For example, the on-board Kawai NV10 engine I bet has a lot of processing going on to make use of the relatively few layers(mixing, splicing, looping, etc) and even with all the trickery one can clearly hear the limited velocity expression when playing it. (Not to knock this piano, it's not bad compared to the competition)
However, with such a thoroughly sampled piano as the VSL Steinway D(I believe they quote 4000 samples per key!), my expectation is that if I select a single mic perspective and happen to exactly request an isolated key in a velocity/configuration that was sampled then pretty much the original sample would be played. I'm not sure why it would need to be dramatically altered if I happen to ask for a sound that was sampled exactly.
I can undersand some EQ'ing to compensate for the microphone response, and layering of release samples and of resonance modelling/pedal noise but otherwise I would hope that what I hear is almost exactly what the selected microphone heard in the room...
Anyway this is sort of educated specilation. I don't really know how VSL does it so interesting discussion!