Note that lack of motion blur is trivially simple to tweak in post-fx to taste.
Not really. I think the general idea is to estimate a motion vector for every pixel or shape, which is actually very close to the same problem as smoothly interpolating between frames in the first place. It's easier to get away with imperfections, though in cases where, for instance, the background is moving relative to the central subject or vice versa, there's no general purpose solution, and in the latter case, there's no solution period, since the moving subjects blur will allow areas of the background to bleed through that may not actually be present in the recorded data.
Obviously there are cases where it's easy to simulate; it's just that there are plenty of cases where there's not just a simple trick you can do.