I wanted to give my take on this example.
So most of the time talking to the Troll King looks like something is going to be at stake: if that's not the case then we (that is, GM and players) can free-narration through it until we get to something that does involve stakes.
Assuming, then, that there is something at stake in the conversation with the Troll King, like 5ekyu I call for a check. In my case this is not so much connected to "player vs character", but to the principle of "say 'yes' or roll the dice", which Luke Crane (BW Gold, p 72) glosses this way:
Unless there is something at stake in the story you have created, don't bother with the dice. Keep moving, keep describing, keep roleplaying. But as soon as a character wants something that he doesn't have, needs to know something he doesn't know, covets something that someone else has, roll the dice. . . . When there is conflict, roll the dice. There is no social agreement for the resolution of conflict in this game. Roll the dice and let the obstacle system guide the outcome. Success or failure doesn’t really matter. So long as the intent of the task is clearly stated, the story is going somewhere.
So the check is "forced onto the player"
in virtue of him/her having pushed play to the point of conflict. (Which of course is what the GM is trying to achieve!, by applying pressure on the players via their PCs.)
I don't generally allow the player to back out: the player should already have a sense of difficulty when getting into the situation and declaring the action, either in virtue of familiarity with the system (this is how 4e andMHRP/Cortex+ Heroic work) or in virtue of a good sense of the state of the fiction (this is how BW and Prince Valiant work, and is also my sense of how most people run 5e).
I think it's important for consequences to be clear - and I don't think of this through the lens of character knowledge but player knowledge - because the player needs to have a sense of what sort of resources to throw at the check (which depending on system and circumstances could be anything from fate/inspiration points, to equipment, to spells and potions, to . . .). Luke Crane says the following about making consequences clear (BW Gold, p 32):
When a player sets out a task for his character and states his intent, it is the GM’s job to inform him of the consequences of failure before the dice are rolled.
"If you fail this…" should often be heard at the table. Let the players know the consequences of their actions. Failure is not the end of the line,
but it is complication that pushes the story in another direction.
Once that is said, everyone knows what's at stake and play can continue smoothly no matter what the result of the roll is.
However, in his subsequently-published book of GMing advice (The Adventure Burner) Crane says that, in his own game, rather than stating the consequences expressly he often relies on context - of the fiction, of the mood at the table, etc - to make them implicit. When I'm GMing, I alternate between express and implicit consequences depending on inclination and whim. But again, for me this has a different motivation from that which 5ekyu states. I'm not worries about player vs character knowledge, and so even if consequence is implicit it will be implicit to the player as well as the GM - there won't be "hidden" bits of the fiction that suddenly emerge into the action on the basis of a failure. It's about pacing and narrative continuity and not weakening emotional intensity with needless explanation.
In a style in which players are "forced" to make checks, and so failure is going to happen from time-to-time (in a system with mechanics that are generous to players) or quite often (in a system like Burning Wheel that is fairly brutal on the players), I regard establishing failures which (i) honour what was at stake, and (ii) follow impeccably from the fiction, and (iii) that are fair to the player and not an excessive hosing, as probably one of the most important demands on the GM. It's very different from a "skilled play" paradigm where it's quite fair that failure at least sometimes be hosing, because skilled players will avoid failure by avoiding checks.
This doesn't seem right. Maybe I've misunderstood 5ekyu's example, but as at least as I'm imagining it either (i) the player has chosen to have his/her PC seek an audience with the Troll King; or (ii) some prior failure has meant that the player's desire that his/her PC
avoid such an audience has not been realised.
if it's (i), well what did the player expect? If you're going to have an audience with a king, it seems likely your interpersonal ability will be tested. If it's (ii), well the player has the full sweep of his/her ingenuity to draw on - anything from blowing up the Troll King, to talking to him, to conjuring an illusion to escape behind, to abasing him-/herself before the Troll King and promising fealty!
Neither pathway to the situation involves the player being boxed in.