This is strange to me. I don't see what the tension is supposed to be.
It's a principle of Apocalypse World to always say what your prep demands. Prep is binding on the GM.
But there is no process, in AW, whereby prep can produce a "negation", by the GM's reference to it, of a player's declared action. I don't think there is in Dungeon World either. As the AW rulebook says, the purpose of prep is to give the GM interesting things to say when the rules call upon them to say things. I would expect DW to work the same way.
I have always understood these sections to be in tension because, as "play to find out what happens" is presented, there should be almost no prep at all, and certainly nothing like knowing very specifically that there is a demon on the second floor of the dungeon who might know about the players. I have had folks tell me, point-blank, that Dungeon World is supposed to be, at least practically if not theoretically,
truly "no-myth", where there isn't any myth, at all, whatsoever, only and exclusively that which is explicitly established in play, and nothing else: hence,
play to find out what happens.
E.g.: "This is how you
play to find out what happens. You’re sharing in the fun of finding out how the characters react to and change the world you’re portraying. You’re all participants in a great adventure that’s unfolding. So really, don’t plan too hard. The rules of the game will fight you. It’s fun to see how things unfold, trust us."
This comes across to me as extremely strong, as advocating avoidance of preparation as much as humanly possible--do only the absolute bare minimum and nothing more. But then, as quoted above, the "exploit your prep" description is almost entirely in the opposite direction, entailing that you already know the whole layout of a dungeon before the party goes there, that you know exactly what is inside (and whether it could turn its baleful eye/s to the PCs), etc.
I've resolved this myself by, more or less, only preparing when I know I need to and trying to keep that preparation very, very light. It's forced me to become much better at improvisation than I was before, I can tell you that much.
So suppose a GM comes up with a brilliant idea for a clue. Then writes a crime that will yield that clue.
The crime was causally downstream of the clue. So by your lights that's not a mystery.
But for all you know, every CoC module ever was written in the way I just described!
Both the clue and the crime precede any possibility of the actual mystery-solvers doing the mystery-solving. The person authoring the story cannot solve that mystery, for the same reason that (for example) a suspense-thriller author cannot actually get the feeling of suspense from their own work, because...they know everything that is going on in the work! They might come up with an idea for an incredible tension-raising scene, and from that tension-raising scene produce a whole dang suspense-thriller novel, but they won't
personally be feeling that thing. The future audience, to whom
both the clue/scene and the crime/suspense are causally upstream,
can solve the mystery/experience the suspense. You can't throw a surprise party for yourself, because you already know about the surprise.
Hence why I referenced the Clue/Cluedo method earlier. Nothing to do with pre-authorship there. Nobody at that table is meant to know the truth--all they have are their own cards and what they can extract from their fellow players over time. But the evidence, and the perpetrator/weapon/location to which that evidence points, is causally upstream of the investigation.
If the chain of events goes...
We investigate -> We establish (by rules, procedures, etc.) things that are true -> If successful, we determine the perpetrator (etc.)
Then I don't see how one can argue that it isn't that establishment (by procedure, rules, etc.) which caused the perpetrator to be one suspect and not another. If you are
causing the solution, you aren't
solving it, you're creating it. What is required, from my perspective, is:
An event occurs -> We investigate -> By our examinations, we learn what already was true -> If we succeed, we determine the perpetrator (etc.)
That's the causal process of solving a mystery. The thing above is a causal process for playing characters who are mystery-solvers, but it doesn't actually involve the personal act of solving a mystery oneself.