But these seem to be simulationist considerations.
Well, I don't know what you mean by "simulationist", but Torchbearer is relatively close to Burning Wheel in its core resolution engine, and both systems use "objective" difficulties that reflect how hard something is to achieve. (This contrast with, say, Apocalypse World.)
So if the cooking skill value doesn't measure how good the character is at cooking, what does it measure?
It does measure how good the character is at cooking. No one has said otherwise.
Why is the difficulty based on how hard the food is make
Because that is how the game is designed? When you (player, playing your PC) try to tackle a harder task, you (player, playing your PC) are less likely to get what you want.
This design feature interacts with rules around PC advancement (while BW and Torchbearer have different rules for advancement, both in practice require the player to make rolls against a variety of difficulties), and with rules around the use/expenditure of player-side resources (fate and persona points, traits, etc).
Like macarons are not harder to make than chocolate chip cookies because baking them makes you more prone to ninja attacks!
You seem to keep projecting your own narrative discontinuities onto other's play.
it is definitely something everybody at the table needs to be onboard with. Otherwise it will easily leed to frustration.
This seems to be a truism about any game. I mean, if I'm not on board with the railroading that is common in much RPGing, I will find that frustrating!
(I personally tend to favour the traditional approach of having test and outcome directly related to each other, but allowing them to be disconnected can make an interesting experience
There is no "disconnection" between setting out to prepare food while camping, and having that process interrupted. They are connected by time, space, the various signs and cues that are created by the process of cooking, and so on.
The roll doesn’t determine “I fail at cooking”. I’m not sure how this could be made clearer. That’s not the result of the failed check. The failed check determines that something happens as a result of the attempt to cook. In this case, the unique nature of the food and the lack of gear meant the cooking took much longer, which allowed the camp to be discovered by bandits.
It could also be the choice about whether to start a long cooking process that involves visible smoke when you know there are bandits about.
These too.
Why would the stakes not be known to the players? I’m not sure how
@pemerton habdled it in the example he offered, but I wouldn’t at all be surprised to learn that the stakes were discussed in some way beforehand.
Torchbearer generally relies on stakes being implicit.
The core rule is presented, fairly crisply, on p 36 of the Scholar's Guide:
The game master decides if there’s a twist or a condition when a player fails a roll. The game master describes the resulting scene. How does the character foul up and get tied in knots? How do they barely succeed? What unforeseen obstacle crops up and disrupts their efforts.
Page 96 adds additional commentary on "Camp Twists":
During camp, the players drive the action. The game master does not present new problems unless the players fail a roll for an activity better undertaken in the adventure phase. If that happens, the game master may introduce a twist as the result of a failed roll. Said twist could be severe enough to cause camp to break prematurely.
And p 223 comments on "Adventures in Camp":
The camp phase is driven by the players. Don’t make an adventure out of it. Let the players spend their checks. Give conditions for most failure results; use any twists generated to build into the adventure phase.
In the episode of play I referred to, the PCs were already pretty heavy with conditions, and the Cook test was the final check spent in the camp. I had to make a decision about the consequence of failure: as p 213 note, "The most primary mechanism of judgement is when to apply a twist or condition to the result of a failed test. That is a serious decision in the context of the game."
I decided that an "unforeseen obstacle crops up and disrupts their efforts" that would "build into the adventure phase", namely, that bandits associated with the moathouse the PCs had spotted, who had formerly been allied with the Dire Wolf the PCs had captured and tamed, turned up at their camp.
The players were disappointed, in the usual way, to have failed the Cook test. But they were not shocked or outraged at the narration of bandits. In fact, as I posted in the actual play thread, the player of the PC cook commented that:
I really liked how everything flowed into the next thing: frogs to feed the Wolf, the Wolf helping fight the bandits, the bandits in the moathouse.