This weekend I played a flash fantasy adventure game. Its content runs through nine levels, but it has a Diablo-style level generator that provides random maps with scaled up enemies, presumably endlessly. Many games have a version of that: levels keep rising with pure procedural content. Infinite levels show an issue common to many games: competing scaling of character power and enemy difficulty.
If power scales more quickly than difficulty, the game becomes trivial. If power scales more slowly than difficulty, the game becomes impossible. If power scales exactly the same, the game becomes very boring as you are doing the exact same thing for potentially infinite levels (“I hit for 25% damage yet again!”)
Any scaling system that runs sufficiently long will be dominated by the subsystem that scales the best. A 1% difference in scaling rate becomes very significant. You may not have noticed it in beta testing: 1.01^30=1.35, so a 35% shift over 30 levels. It is +64% at 50, +170% at 100, +345% at 150 +632% at 200. Wow, those numbers really started jumping, didn’t they? Behold, the power of compound interest. And there are games like Asheron’s Call that scale that far. If melee improves slightly faster than archery, melee will dominate the late game; if one weapon type improves slightly faster, it will be the only one worth taking.
D&D players are very familiar with how this happens in only 20 levels. A level 1 Wizard is a sleep spell in a dress, while a level 1 Fighter can take out quite a few goblins. A level 20 Wizard is a nuke-tossing god who alters reality at a whim, while a level 20 Fighter can take out quite a few larger monsters.