Well, given that there are, as you say, infinite potential variables in an RPG (lets not debate whether this is a universal truth about RPGs or not) then what purpose do the choices of framing by the GM in a game similar to DW really accomplish then? Lets look to the obvious place, which is the stated agenda of the game! Now, that will differ from game to game, so we cannot really make generalized statements about it, except to say that presumably good GMing serves that agenda, and bad GMing does not. There may also be other criteria we want to measure as well. So, I don't care about infinities in that sense.Absolutely. We have - roll dice for each decision - as a rational exit from the dilemma that I outlined.
We're likely to find some clear water between our views here, as for me all TTRPG worlds have a vast number (literally infinite) unconstrained variables even if they are not realistically detailed. It might not be worth our debating that, but just accepting that we have differing views.*
From descriptions of play and watching videos, I can see that DW is revealing in the way it manages and is successful with - roll for each decision. In terms of unconstrained variables, it relocates them (e.g. to 6s or less) and declares boundaries to the decision space (so that it is vast, but less vastly diverse - I recommend reading Borges Library of Babel to grok that).**
Still, one can see that whereas in an example a wand bounced down a shaft, there were a million (again, genuinely infinite) things that could have happened to that wand: the precise outcome wasn't defined and wasn't rolled for. Most choices might have amounted to different flavours of the same thing - e.g. 'I can't use the wand', or, 'I can use the wand if'. But then, there are so many ways that "if" could have gone - if I wait, if I do X, if I pay Y, if I roll and succeed at Z, etc - and the knock-ons from there.... a vast cascade.
* So I am saying that the intent of - "reduce the number of variables, by reducing the realism of the fiction and stepping up the number of recognisable tropes/conventions/stereotypes" - rationally only changes one kind of infinity into another kind of infinity. That is true unless you reduce to a stochastic or deterministic mechanic with properly bounded inputs, functions, and outputs. The second infinity is more useful (see below), but one thing it is absolutely not doing is removing DM-fiat. Only refining and relocating it. That is worth doing, but it is not the same as removing DM-fiat. Rather, we are making statements about better and worse kinds of fiat for given purposes... and there is nothing wrong with that! It is helpful to do.
** Making a space less vastly diverse is extremely valuable. It becomes a helpful subset of infinity (albeit also infinite). Say we have a principle - choose only numbers divisible by 7. There are an infinite number of them, but still fewer than there are numbers overall, and if what we care about is divisibility by 7, then we have done something useful. The picture is not quite so clear with TTRPG principles of course, as we see in debates as to what principles going by the same name include and exclude.
EDITED My view is more open than was captured in my words. Also, removed some text that might not have helped my explanation.
When I talk about infinities, I think what I'm talking about is how infinite amounts of indeterminate implicit state of the game world undermines any possible attempt to generate arguments about what "logically would happen next." Take the 'bribe the guard' scenario. There is literally no way for a player to know what might come about if he approaches a guard and tries to bribe him. Maybe the GM has some notes which amount to stating a result of such an attempt (presumably it is foreseeable). Maybe it is just stated in terms of probabilities, modulated by some character trait or varying depending on easily classified game state such as the PC's approach, his race, social class, appearance, size of bribe, etc. This is fine, but it also may be that all this is sort of eyeballed on the fly. That's OK too, but the player really cannot judge, except by simply querying the GM's imagined world state. If he is going to simply work from what is likely to succeed he will soon run into problems. Maybe his PC is a dwarf and all the guards hate dwarves, how would anyone know? Even the GM may not know, so how are tactics even possible here?
This is all why I have long favored approaches similar to DW. You have stated explicit agenda and derived principles, coupled with a system which generates story outcomes based on that, or at least suggests to the GM when and how elements might be introduced to the game and mechanisms for doing so, pacing, etc. Clear agenda, everyone knows what game they are playing, and clear process which supports that agenda. Sure, there are 'infinities' and plenty of GM 'fiat' in DW. But like the example @Manbearcat posted that you paraphrased above, the GM is simply applying his process there. The Wizard rolls a failure/consequence, and he puts stress on the situation by having a complication, the wand shoots out of her hand. Now she goes after it. The player here articulates an in-fiction tactical reason for that move, but it also serves the game by upping the ante. All her cards are now on the table, character death is plainly being wagered at this point. It is consistently in-character also, which is another requirement besides plot-wise fictional consistency. This ticks all the boxes of DW.