How Long Should an RPG Be?

Inspired by Ars Technica's tablet PC article, Tales of the Rampant Coyote has written a piece about the appropriate length for a role-playing game.
When I was at Infogrammes (now Atari) , the company head Bruno Bonnell told us that a study had shown what many of us suspected - the majority of players never "finish" their games. They would eventually give up and move on to the next game. Bonnell's contention was that we were wasting half our development efforts if the players were only playing half the game. I don't think the math necessarily applies - you reuse a LOT of code and content in later levels that is needed throughout the game.

While he wasn't speaking specifically of RPGs at the time, but I've little doubt that most copies of RPGs in the 1990s never accessed the "ending sequence" file on the hard drive. Players tend to play until they grow bored or frustrated, and then quit. Those aren't reactions any developer wants to his or her game. But even the most sadistic game designer really wants and expects players to actually see the endgame.

...

I think most players would really rather play a 20-hour game full of extremely awesome, high-quality, engaging content than a 100-hour game made up primarily of "meh." But I do wonder if there isn't a target value of an optimum average number of hours in an RPG before players start growing weary of even the most well-crafted storyline. I think the law of diminishing returns does apply somewhere where even the most devoted players may waver in their attention.
I'll take a great 100-hour game over five average 20-hour games any day. And maybe I'm more patient than most people, but I've probably seen the "ending sequence" to 90% of the RPGs I've played since the early 80's... even when they're pathetically awful like the exit-to-DOS-with-a-paragraph-blurb ending in the original Eye of the Beholder.