What counts as good or bad faith depends (doesn't it?) on prior commitments and understandings.
Yes. I think some understandings are baked into the games, and others are baked into people's preferences regarding games and/or playstyles. I think it's plausible you and I don't have radically different ideas about good faith play/GMing, though we have, I suspect, markedly different preferences as regard playstyle and systems. I'm pretty sure neither of us is entirely wrong;
If I sit down to play the Dragonlance modules, then presumably it's agreed that Kitiara can't be killed by a few lucky bowshots (to get the damage high enough let's suppose they're from an Unearthed Arcana bow specialist at point blank range) early in the module series.
Aren't the Dragonlance modules a pretty notorious railroad? I've never played them (or, for that matter, read the books) but it's my impression that the modules are pretty specifically about giving players a chance to experience the books as a D&D campaign. Given my feelings about books and TRPG play being very different types and experiences of fiction, that seems like a very, very bad idea.
That said, I agree that killing one of the main characters of the books, early in a campaign through those modules, would almost certainly be against the players' expectations (and probably the DM's, too).
If I sit down to play a standard game of Burning Wheel, then the GM deciding that there are no secret doors and thus not allowing the action to be resolved in the normal way is acting in bad faith.
That is consistent with your descriptions of play. You mentioned early that plausibility-checking in TRPGs is interesting, and I agree; how is it handled in BW--who, if anyone, could say it was implausible for there to be a secret door in that wall? Also, if the GM ended up narrating (IIRC, because the player failed at the action-resolution check) I believe their options would include "there's not a secret door" as well as "there's a secret door and you can't open it" or "there's a secret door and it leads to something even worse," yes?
I don't see Orcs, ice walls or secret doors as being a priori different as far as unilateral GM decision-making is concerned.
I think there's a case for all three being in the realm of fiction-framing. Which, in D&D, means they're placed by the DM, yes.