It seems the major downside of recording actual audio is the limited moddability in making a flawed human recording sound more perfect, while the major downside of MIDI recording is the time and effort necessary to make perfect flat synthesized sounds, sound more human. Both are cons; each are from one extreme.
Are MIDIs/VSTs less appropriate for more expressive / subtle / nuanced/ classical compositions? For example, even with Vienna Symphonic Library's VST (one of the better libraries?), it seems it would take an inordinate amount of time and effort to reproduce the tone color and rubato of human playing (see link below). But this (skilled) musician can do it in minutes.
How difficult is it to mimic the first 60 seconds of this video using VSTs and all the mixing effects at your disposal? The cello play sounds pretty detailed.