Bike porn

All that and no one's called out the now-dismissed-by-science q-ring.

See the timetrialling thread

Surely in a track sprint, as little backlash as possible = a good thing.
Right?

Who knows? If zero backlash were an unambiguously good thing, we wouldn't have cush drives. A quick and dirty Fermi Estimate suggests that the pedal position might lead the wheel position by about 1° of crank rotation at peak torque from chain elongation alone. If we look at the Q-Ring problem above, it's pretty clear that it's all but impossible to find any biomechanical/physiological effect with such a small variation in the effective gain ratio at different crank phases (Q-Rings obviously provide a much bigger variation), and the hysteresis problem is vanishingly small too.

By letting the pedal effectively make a run for it when the rider applies maximum force, the actual peak force on the body is slightly reduced (same argument as a cush drive reducing peak loading in an automotive drive train). Although this is a tiny effect, an argumentative person might hypothesise that protecting the muscles from extreme overload in this way could reduce the amount of micro-rupture in the muscle fibres, leaving more fibres available for work later in the race. This would be an argument for having more elastic backlash on your race bike than on your training bike :-)