I like this, but given my perception of your dming style from posts of yours, I would imagine this was the PC's idea or am I wrong on this assumption?
It was the player's idea. At the start of the campaign his wizard had, as backstory, former cultist of the Raven Queen. Over the course of play (15 levels, approximately 3 years in real life) his PC became increasingly connected to a range of gods, and was a multi-class invoker. He decided the PC would make more sense rebuilt as a deva invoker multi-classed to wizard. (In 4e multi-classing is a feat that gives you modest abilities associated with the multi-class.) The actual transition events we worked out together - eg I think I'm the one who suggested that his imp familiar would be a spy for Levistus, placed there on instructions from Bane.
If this idea was entirely hatched up by the DM and he placed this in the PCs backstory without consultation with the PC, would you have an issue with it?
A big issue, yes.
I don't mind the GM springing surprises on the player, even including surprises about PC backstory. But the player has to have asked for it (either expressly or implicitly) so that it fits with where the player is taking the PC. Eg with the same PC, back when he was still a human, I placed an encounter with his mother as a prisoner in a goblin cavern complex. This made sense given the player-author backstory (family scattered when they fled their home city as refugees following a humanoid assault). But I decided to actually introduce the PC's mother into the game.
In my view there can be no hard-and-fast rules here, because different players have different degrees of tolerance for GM surprises, and different degrees of desire to have the GM pick up on their backstory and run with it. That said, I wouldn't want to run a game in which every player is a turtle who retreats into his/her shell at the merest hint of GM playing with backstory. That doesn't sound like very much fun to me.
I'm not a fan of the latter idea from the Adventure Burner. It would suggest that PCs can expend all their dallies and specials on the little fights before the big bad and as DM you would have to scale down the big bad for it to be a suitable challenge for the PCs when they do encounter him. This is where it seems the world is not intuitive/natural. As DM the Adventure Burner seems to suggest deliberately manipulating the big bad for a fair challenge as opposed to preparing the monster and so what happens during the adventure/session does not affect your big bad unless it was part of the in-game fiction influenced by the PCs. I'm more of a fan of let the chips fall where they fall.
I think the Adventure Burner advice is a bit conflicted. Burning Wheel is based - and very strongly based - around "objective" DCs. But by D&D standards it has relatively modest scaling, in part to make those objective DCs workable in a practical sense. It also has a range of devices - both mechanical and GMing techniques - to handle PC failure, which is expected to occur much more frequently than is the case in D&D. So it's very much a "chips fall where they fall" system.
The main worry that motivates the Adventure Burner advice is that if you stat up the NPC too early, the actual resolution of what should be an epic confrontation will fall flat, for rocket-tag type reasons (on one or the other sides). Once you know what the PCs will look like, mechanically, when they meet the big bad (BW doesn't have levels, but rather skill numbers) then you can set the big bad. There is nothing really analogous in BW to expending dailies or specials, so there's no particular need to modulate difficulty to reflect the PC status when they encounter the NPC. (BW does have fate points and similar things, but the game is designed so that the players should be accruing plenty of them as they come into their final confrontation - if they are not, then the gameplay has misfired in much bigger ways than are involved in statting up an NPC.)
Because D&D doesn't have modest scaling (even 5e has very rapidly scaling hp), and doesn't have the devices BW uses to handle PC failure, and is chock full of dailies and similar "specials", I don't think the BW advice is immediately applicable. At best it can be suggestive.
Finally, on this point, I don't get the bit about intuitive/natural. The NPC has no stats until the GM writes them. Provided the stats that the GM writes up are consistent with what has actually been established, in prior play, about the NPC, how is one set of stats more or less intuitive/natural than another? Eg if all we know about the NPC wizard is that (i) she has lots of followers, and (ii) she once cast a Gate spell, then either of the following strikes me as equally intuitive/natural:
(A) She is a high level wizard who can memorise and cast Gate and wields a Rod of Rulership;
(B) She is a modest level wizard who has a high CHA, as well as an empty parchment in her library that once was a scroll with a Gate spell on it.
I don't see how it makes any difference to the verisimilitude of the play experience if the GM decides this in advance, or on the spot.
