I think we tend to exaggerate the extent to which GMs have a more complete view of play than other players.
<snip>
I prepare on a session by session basis. I use the established setting to try to put pressure on players to make interesting decisions, but I have a relatively light agenda. When play is proceeding on a solid clip and players are engaged, my job is easy. I do minimal prep with an emphasis on ensuring the initial moments of play drive towards an exciting narrative, but am otherwise hands off. Setting and situation creation decisions are made with an eye towards play and can often occur in the middle of the session.
All of that makes sense to me.
In my most recent session, several of the PCs got sucked into an Abyssal Breach. According to that stats of the Breach (with which the players were familiar, via PC knowledge checks), a character who is sucked in is "removed from play" - which, in 4e, means no actions and no line of effect to anything. But with each of the PCs sucked into the breach, the player made a case for having his/her PC do something. The primordial-worshipping, chaos sorcerer Demonskin Adept - who had been hoping to second wind on the turn that was lost by being removed from play - sent out a call to any chaotic power that might hear him, and was answered (as it turned out, not by his patron Chan, Queen of good air elementals, but by Yan-c-bin, Prince of evil air elementals) and hence got his second wind. And the Sceptre of Law (= Rod of 7 Parts), Erathis-worshipping Sage of Ages, who had been planning to bring down the breach from the outside, started making knowledge and skill checks to bring down the breach from the inside (and made skill checks to avoid being spat out of the breach when its "removed from play" effect normally would end) - with the help, it turned out, of the paladin of the Raven Queen who also got sucked into the breach.
The presence in the game of these various story elements - the PC backgrounds and loyalties, the Abyssal Breach, the conflict between Chan and Yan-c-bin - were all established. Likewise the basic mechanical framework. But how it is all going to play out; what steps the PCs will take; the idea (for instance) that one of the PCs might get sucked into the breach and in doing so receive a boon from his patron's mortal enemey; all that, including the elaboration of backstory that it involves (does Yan-c-bin know and/or care about Chan's disciple? what happens when you go inside an Abyssal Breach?), is worked out during play.
The above example deals with somewhat micro-level details of action resolution, although they are tightly connected to broader concerns of the campaign. But I find the same approach can be applied to more macro-details as well, like trips from A to B, and adjusting what is expected to happen at B in light of what the players want their PCs to do, and to be involved with.
If the DM has a scene in mind that none of the players have any investment in, and can't be engaged with until the DM forces the players to play out that scene, then the DM is doing a very, very poor job. If the only person at the table with any investment in the next scene is the DM, then the DM has completely failed to convey any important information to the players and has failed to hook the players in any significant way.
I would add to this that the GM has a different role from the players in most approaches to D&D.
If the GM is a referee of Gygaxian/Pulsipherian dungeon crawling, then the GM must (i) set up an exciting, engaging, perhaps wacky dungeon and associated environment, but (ii) not let his/her desire for excitement/wackiness affect his/her refereeing in any sort of unfair way. The GM has to be neutral, and not change the parameters of the situation on the fly, or else the players can't make meaningful choices, use their divination spells, take advantage of prelminary mapping and scouting, etc.
If the GM is a referee of a worldbuilding, sandbox game then the GM must (i) build an interesting, "realistic" (in the appropriate sense) gameworld, but (ii) must let the players make choices for their PCs that are themselves "real" within that gameworld, and that might change the world. The GM in this sort of game who constantly blocks player choices, or who takes steps (eg via NPCs, gods etc) to ensure that player choices never really make any difference to the GM's world, is in my view not being fair to the basic premise of thise sort of game.
If the GM is running a scene-framing game that is player driven in the sort of way that Hussar is describing, then the GM still has a lot of power - aspects of world building and theme-selection may be shared, but the GM still probably has the lion's share of authority over backstory, and over the narration of complications. But in order to respect the players choices about theme, PC backstory, thematic focus etc, the GM is obliged not to run roughshod over those by constantly framing situations and narrating compications that speak to none of that. Which is, I think, Hussar's point.
What is the role of the players in a game in which the GM has authority over backstory, and world-building, and scene-framing, and narration of complications, to the extent where these powers may be exercised entirely as the GM thinks would make for an interesting story, with no expectation that these powers will be exercised with regard to the players' evinced preferences (as evinced either at start-up, or in play)? As far as I can tell, all the players get to do is narrate their PCs' responses to ingame situations, in effect adding colour to the GM's story but not much else. That seems to be a more common approach to RPGing than one might suppose a priori, but I take it as obvious that it's not the only viable one.