Tris: I like the idea of keeping the pressure on by being unpredictable. But AFAICT, going for 2 works somewhat less than half the time. If there were a 50% or better likelihood of scoring on a 2-point try, you bet Spurrier would be going for two all the time, even if nobody else did. Look at the way he’s willing to take a shot on 4th down. (Hijack: I liked his 4th-and-5 gamble at the Rams’ 34 the other day, even though it failed. It was too long for a FG, and the likely gain from a punt was only 14 yards. What was there to lose?)
Lorenzo: you got a cite on the .33 chance of success on going for 2?
John Carter: even though it didn’t cost them the game, my favorite instance of this was USC-Notre Dame, 1974. Notre Dame took an early 24-0 lead (3 TDs, 1 FG, like you’d expect). When the Trojans finally started scoring, McKay had them go for 2 after the first two scores. Failed both times. Finally went for 1 after the 3rd TD that made it 24-19 - but it would have been 24-21 if he’d settled for the PAT all along.
Fortunately it became moot: the next score was also a USC TD, and they went for 2 and made it: 27-24. They ultimately scored four more TDs to run it up to 55-24. As the saying goes, you can look it up. 
There’s one NFL game, from before the 2-pointer days, that I wish the option had been available for. Redskins-Cowboys, last game of the 1979 season, in Dallas, a playoff spot in the balance. A real seesaw of a game: Redskins ahead 17-0 early, Staubach brings 'em back to lead 21-17, the Redskins go back up 34-21 midway in the fourth quarter.
Going for 2 to make it 35-21 would have been awfully sweet. Because Staubach brought the Cowboys back for 2 TDs in the last 4 minutes to make it 35-34, Cowboys, which was how it ended.
That’s the sort of situation that’s tailor-made for the 2-point conversion.