It is interesting that we think of chains made of steel as stretching and belts made of rubber as non-stretching. Of course, metal does stretch, but it is more likely the wear in all the rollers that is lengthening the chain. I would also note that chains seem to have a much more difficult time keeping the timing accurate, due to the wear. Belts are somewhere in the +/- one degree area where chains are in the +/- four degree area as I think Mercedes found out.
If you ask: "Is a cush-drive necessary" the real question is "necessary for what?" Racing organizations like them for the safety aspects. They tend not to hurt riders when they break and they tend not to to hurt the bike, which means less oil on the track. If you run a Vincent with a MK2 cam the power is between 4,000 and 6,000 rpm. The power pulses are not exactly smooth, but the street rider may never get above 4,000 rpm. In this range the pulses are quite pronounced.
I do not run a cush drive anywhere on the racer, but I am spinning it fast all the time, even on the starting line. If I had power related reliability issues I would consider adding the complexity of a cush drive as a worthwhile addition. I think on the street, cush drives have proven quite useful or we would not see so many modern bikes using them. My Egli always had one in the Campagnolo rear hub. It is something I would design in to a new Egli as Glen has done. I think there is a lot of "up" side and very little "down" side to running a cush drive. It seems to be a good solution for low rpm stop and go riding.
David