My first thought here is to ask where the problem lies in this scenario. Stuart calls the situation a slippery slope, but as described I see a basic GMing technique. Only the dimmest of players will fail to go beyond the scope of the GM's initial creations. My old HERO System group called it the inevitable "X+1", where the GM plans for X number of contingencies and the players come up with at least X+1 responses to the situation. Rolling with these sorts of punches is something every GM does. Nothing about the dungeon makes that different.
This slippery slope is something that I suspect occurs across multiple games rather than in a single game.
Game 1: Players investigate. GM makes stuff up for them to find.
Game 2: GM makes up stuff ahead of time for players to find. Players ask questions/do things GM didn't anticipate.
Game 3: GM tries to create more coherent setting so that making stuff up will be easier...
Does that make sense?
Let's give a concrete example. The party has discovered a treasure that by random die roll happened to include a large number of potions. One of the players gets curious about this and innocently asks "Are all these potions in the same kind of bottle?" The DM has nothing in the key one way or the other, but he sees a golden opportunity. "Yeah, they're all shaped like those wide-bottomed science flasks, but made out of green glass. Now that you mention it, the potion of heroism from the bugbear lair is in the same kind of bottle." While the players spend 10 minutes discussing the ramifications of this, the DM sneakily writes down "new sublevel 4a, lair of the Mad Alchemist, include Ogre Mage". The next time the players interrogate a prisoner, they might ask the poor critter about potions. The DM is ready now. He tells the PCs that a strange blue giant carries a sack full of them around the dungeon, selling them to monsters for a hefty fee.
To me, this sort of operation is part of the ongoing process of good megadungeon development. Any novice DM can throw some dice on a random stocking chart and sketch some tunnels on a piece of graph paper. To me, that's just step one. A good dungeon makes something out of those dice rolls. And just as importantly, a good megadungeon is never completely done. Adding new sublevels, rooms behind secret doors, and other expansions is an ongoing venture.
Going back to the succubus in room 23 and the gold dragon in room 24, I think S. John is right on the money when he dismisses these sorts of incongruities as basic axioms of the genre. A DM can take that sort of situation and spin a plot out of it. Or he can simply use it as further evidence that the dungeon does not always operate under the same rules as the rational surface. Stuart is right on the money to spot a contradiction here. But I choose to work with that contradiction, even to embrace it. A good dungeon adventure is like a journey to a nightmare realm, where bravery and smarts are your only tools and sometimes they just don't work. Is that unfair to the strategic player, whose bestlaid plans can be screwed up by one bad wandering monster roll? Yes, it most certainly is unfair. And I wouldn't have it any other way. Dungeoneering is not for the faint of heart.