I don't entirely agree. One thing that is always stressed in these sandbox discussions is the notion of meaningful choices and meaningful consequences. The GM knew that there was nothing in that drain. This was a pre-written adventure.
In which case, I would let that stand.
Go back and read my post. I suggested the following:
1. The GM thought more about it beforehand.
2. The GM made it a single encounter (the same as not having to play through, say, chewing your food at the inn).
3. If the GM did not know what was down the drain, he could consider what might be there on the fly.
From what I can tell, it seems to me that the GM did not know what was down the drain in the example, but mentioned it as something that should be there.
Thus, the players could choose to go there, but the "meaningful consequence" was perhaps an opportunity cost, but
not an unfun session.
If the GM decided, in that moment, that the architecture of the fortress had changed to accommodate player whim, that sounds like illusionism to me, which IMHO is less compatible with sandbox assumptions.
I would agree. But, consider that the GM might have marked the guardrobe and the drain without any more thought. This is not dissimilar to, in a wilderness scenario, telling the players that they can find an oak tree even though you haven't detailed every tree in the forest.
In fact, I'm rather surprised to hear you say that you would place lairs and encounters where the players feel like going, rather than by some more abstract logic.
Again, note "if the players go somewhere that I haven't fully fleshed out". I don't change work that has been "set" (whether the players have visited it or not), unless it changes as a consequence of the world itself (some orcs move in, an NPC party goes through the area, etc.). But there is nothing wrong, IMHO, with having materials ready to help in fleshing out areas on the fly (if needed). In addition, if one uses wandering encounters (and I do), there is no problem whatsoever in including minor lairs within the framework of those encounters, especially if these lairs are of a temporary nature (orcs with a campfire, say).
Obviously, the more prep work done, the less one has to rely upon "winging it", but one should always be prepared to "wing it" if one has to.
EDIT: Just re-read that post, and I can see how it might imply change during play. That was not my intention at all. My intention is:
1. A GM thinking in sandbox-mode while creating an adventure is more likely to think in terms of total environment, and therefore less likely to make the drain dull to begin with. This is while game prep is taking place.
2. A GM who made the drain dull, and knows there is nothing there, can treat it as a single encounter that takes no more than 5-10 minutes (on average) at the game table.
3. A GM who has not even considered what is down the drain should be prepared to "wing it" and do what he would have done while prepping, had he thought about it: either put something there, or do not put something there and make it a single encounter. If the players exhibit sound reasoning as to what is there, and the GM has no reasoning of his own, he should consider stealing their reasoning.
RC