dimples, like in a golf ball, work by tripping the boundary layer to become turbulent, rather than laminar. this keeps the flow attached (as there is more energy in the boundary layer) and this reduces Cd. i've noticed that BMW have added some dimples on the stalk for the rear view mirror, probably to reduce noise (from separation of the airflow). as far as applying this to bicycle wheels, i think there are many factors to consider ... reynold's number, etc. you mention yaw effects ... bicycle aerodynamic improvements seem to be very sensitive to these effects (sometimes in cross winds they'll use solid wheel discs, sometimes not ... it seems to be more a personal chioce of the cyclist, than a demonstratable advantage. most likely there are many conflicting issues ... a slight redcution in drag countered by more work to spin the disc offset by more inertia in the disc (which can be good on a straight course and bad on a twisty, turny one).