Curveball Design

Over at Zen of Design, Damion Schubert muses about all the bad endings in games. Of course, as a designer, he focuses on the badly designed endings (as opposed to poorly written endings to the stories). When you think about it, it does seem like an awful lot of great games end with curveball design elements.

• The jumping puzzle at the end of Half-Life
• The jumping puzzle at the end of God of War
• The jumping puzzle at the end of Psychonauts
• The jumping puzzle at... wait a minute...

Ok... so maybe we should just expect a jumping puzzle at the end of every game. Seriously though, the thing that's so jarring about these jumping puzzles is that they come at the end of games that aren't about jumping (except maybe Psychonauts, but I hear that that puzzle is insanely hard compared to the rest of the game). I can imagine that these design choices are made to provide challenge and perhaps urgency to the end of the game, but having to adjust to a new mode of play can kill any momentum that the player has going, which can destroy immersion and cause frustration.

Schubert suggests that twists in the design might be more elegant. Adding a new layer or wrinkle to existing mechanics would challenge the player without breaking immersion or causing unnecessary frustration. Since the player has spent the entire game learning and mastering a set of mechanics, it makes sense to maintain that learning curve (or even reward it) with additional variations rather than completely new challenges. You don't want to make the player feel like a beginner after they've played your game for 12 hours.

That said, is there any time when a complete curveball game mechanic actually works near the end of a game? Any examples (I haven't come up with one yet)?