Yes, i thought of the Kiesse win, but didn't include it as he was in a long break and the attacked his companions with about 3-4 km to go. But you are absolutely right, chain falls off in the crash, took him two or three tries to get it back on, and then still had enough to beat the pack. Granted, the pack had to slow down significantly when they entered the tricky turn that brought down Kiesse in the first place, but just remarkable.
Also +1 to Ygduf's point, though it's not exponential but cubic
. This is why one should go from further out in a tailwind sprint as once you get up to speed, it's damn hard for anyone else behind you (who has to first make up a bike length, viz. go at a higher speed) to overtake you.
Also @hidayanra, i've never pulled off this sort of thing myself. Tried it once with a bit more than a mile to go in a crit and spent five hellish minutes dangling. Got passed with about 300 m to go and ended up 7th.
PS. The examples re: Stybar's win last year, Oscarito's first win, and Voeckler's GP Quebec don't apply. The first two are from small groups, where there's a lot more people watching going on than when you have dedicated sprint trains chasing. One could add Dan Martin's Lombardia win as it's exactly the same
move that Oscarito pulled: attack from the back with about 700-800m to go and catch people off guard (which is perhaps the best card a
non-sprinter has when there are about 5-10 people contesting the win). Voeckler's attack occurred on a hill and thus renders the conversation moot.