I recall seeing someone post frequency response curve or spectrograph or something of the NV-10's output. It definitely does something to match the response curve of an acoustic piano, and I assume the sound engine's output is matched to that. I.e., the speakers are tuned specifically and only to the NV-10's sound engine characteristics.
However, I would imagine that most VSTs are designed to be output via more neutral studio monitors or headphones, so piping that into the NV-10 might result in some unanticipated filtering/adjustment of the sound.
I've used the NV-10's Bluetooth Audio feature to use it as a jukebox/speaker, and it's indeed more than a bit odd. I can't tell what it is, but it really doesn't sound like just a big speaker array, it definitely filters the sound in some way. I expect it's doing the same to VSTs.