The GM's conception of the situation did not change one iota. The situation begins as
PCs must do XYZ in order to get <whatever it is that they are supposed to get> from the Mad Tyrant> And the situation ends up as
PCs not having done XYZ are not getting what they want from the Mad Tyrant who, being a Mad Tyrant, is jailing and executing them.
The GM decided what was at stake - namely, the PCs getting <whatever it is that they are supposed to get>. The GM decided the significance and implications of the NPC being a Mad Tyrant. The players' conception of the situation - that it is about the moral stakes of compromising with, versus opposing, a Mad Tyrant - gets no look in at all, to the extent that the GM, posting, describes that conception of the situation as "murder hoboism".
When I talk about play being a railroad, or play being driven by the GM, I am not supposing that the GM has the players bound and gagged, or that the GM just recites a monologue. I am talking about exactly this sort of thing: the GM frames the situation, decides what is at stake, and establishes the consequences of the players' declared actions entirely by reference to those decisions that they have already taken, so that what the players are doing in play is learning what the GM thinks the situation is about and what should follow from that.
I don't know if this is
exactly what
@soviet had in mind in posting "ultimately in a trad game it's just colour", but I believe that it's at least in the neighbourhood. Likewise I think it's pretty close to being an example of what
@hawkeyefan has in mind in saying that "these choices yield different results as to the content of the fiction. But not to its nature."
I didn't use the phrase "change of direction". I am talking about
what is colour? and "content vs nature of fiction*, building upon my understanding of hawkeyefan's and soviet's posts.
The content of the fiction changes - as, inevitably, it must, if people at the table are declaring actions and resolving them. But the GM's conception of the situation remains absolutely static. There is not even a thought turned to the sorts of possibilities that
@thefutilist mentions - guards refusing to obey the Mad Tryant's orders and joining in the uprising; locals freeing the trapped PCs from the stocks; the Mad Tyrant himself repenting; etc.
I'm pretty confident that this is what
@hawkeyefan had in mind in saying, upthread, that
It also illustrates how
one sort of approach to GM prep is at odds with player-driven play: namely, when the GM's prep produces a completely static conception of
what is at stake and of
what might happen next.