I don't think an endgame stratagy is a bad thing as long as it is 'endgame'. Super powered armageddon spells and ultimate god-like heroes are common through fantasy literature and games, so why not here? As long as someone can't 'win on turn two' in multiplayer, than a well thought out and executed stratagy SHOULD eventually snowball your opponents if they havn't done anything to counter you. That said, I think there should also be many different endgame stratagies to choose from, be it controll a dragon, make a super hero, or cast a super spell, that will give as good as they get, so if two opponents can build up without countering eachother, well, theres a truely epic strugle to be had....
Yeah, the point is, once you've played the game enough, you can usually tell who is going to win the game very early on (or at least whether or not you will be said victor). And what Vordrak is saying (and most of the members of these forums agree based on previous discussions) is that it should be impossible to say who is going to win a game until the game is almost over. That means that your first few choices in the game can't be the overwhelmingly biggest factor in the outcome (which they usually are), and it means that just because you're the biggest and strongest midway through the game you still can't count on winning. I think Elemental is on the right track on making this happen with mechanics like essence, master quests, insanely powerful individual characters (player or otherwise) and intentionally debalancing world-wrecking spells.
Also, this is just as important in singleplayer mode as in multiplayer modes. Yeah in multiplayer modes the snowball effect usually leads to the losing players leaving, but in singleplayer it's not much worse. You get 30 turns in and realize "ok, well I've won but I still need to play another 100 turns to see the victory screen." Ideally, by the time you can say "ok, I've won," the victory screen should be right around the corner.