I want to split the hair one micron further:Well there is an important hair to split:
If you want something to have the character of a _puzzle_ (this includes actual puzzles as well as encounters and mapping challenges and other complex kinds of multi-step secrets) then it helps to have a thing prepped.
It doesn't "have its own reality" but it sure as hell makes it easier to create a puzzle-style challenge and the kinds of challenges related to that.
I think in some puzzle-oriented play (eg classic dungeon exploration in the sort of mould that Gygax sets out in the closing pages of his PHB, that Lewis Pulsipher used to articulate in late-70s White Dwarf, etc), there is an implicit commitment by the GM to have the puzzle elements authored in advance so that the players can then deploy resources (eg ingame time, detection magic, etc) to work out those elements and unravel the puzzle.
This isn't necessarily related to "objective reality" - eg in this style I think it's completely fair to have rooms with an A% of thing X or (100-A)% of thing Y (eg the ogre is in their torturing the kobold, or is down the hall having a nap). And that "reality" won't become a thing until the game is actually played and that door listened at or opened.
But the GM shouldn't be toying with those percentages during play. In his DMG, Gygax sets out this sort of idea of GM's notes as a pre-play commitment in the rules for evasion of dungeon encounters: if the GM's notes say that the monster does or doesn't pursue, then that takes precedence over everything else.
Only if the GM promises to hold the parameters of the puzzle constant can players solve it using the resources that the game gives to them (via their PCs). It's about fairness and the necessary conditions of a certain sort of puzzle-solving. (It's certainly not about creative integrity, at least as far as I can see.)