And there is another aspect of resampling keyboard sounds that no-one has mentioned.

Simply sampling the final output of a complex multisampled instrument consisting of layered samples most of the time will never compare with the original hardware. Sure it may sound ok but it will never have the same dynamics. Layered sampled sets eg consisting of attack samples, resonant samples and maybe even multiple sustain samples plus samples that switch based on velocity can have independant ADSR, filters and velocity response so the sound responds to your playing style can't be captured accurately by simply sampling the overall output.

If I like the way a keyboard sounds and responds to my playing style then I'd always choose the original hardware. Any sampled copy will always be a second rate imitation.