Personally I believe it would work pretty damn well, and be very much preferrable to motion smoothing, as long as the timing of the action in the frame is correct (that moving bicycle needs to be in the right spot, when the screen flashes on), and the frame duty cycle is right for the duration until the next frame, so that the light level does not fluctuate.
Predicting frame times is the hard part, of course, and there is undoubtedly always the risk developers would use it as a crutch, but in both cases no more than we already have with motion smoothing – argueably potentially less so, since the developers can not let the time between refreshes drop too far, before the view starts to flicker.
It would allow for really playing at arbitrary refresh rates like 73, or 94.2 Hz, if that is what one’s hardware can pull for periods of time, without having to wait for the next refresh cycle, and slipping even farther behind.
(I also firmly believe motion blur would work great in VR, if realistic (therein lies the rub) blur could be produced cheaply – no more strobelight laser pointer trails.)