The physics of it wouldn't actually be that hard...2D roller coaster physics is incredibly easy. Moving it into pseudo-3D isn't that much harder, and you can pre-calculate a lot of things because the track is made up of known segments.
Edit: Completely realistic rollercoaster physics is not easy, of course, but RCT physics were pretty basic. Just a accelerations, Gs, and sometimes thresholds for things like coasters flying off sharp turns for the one(s) not attached to the track...