Sessions of campaigns can't exactly follow this model, but because of its limited scope I think this kind of "episode" can be more easily transferred to campaigns than other structures. You take the PCs, give them conflicting incentives or highlight their differences, and then impose a constraint such that no one can be satisfied until all are, and let them get at one another. The main thing to watch out for here in a roleplaying game is setting up a constraint that isn't actually a constraint, like putting the PCs at the bottom of a giant pit when someone has levitate, or thinking that a bunch of NPC guards preventing them from exiting through a doorway will keep the PCs in a room. You can amplify an otherwise crappy constraint by offering incentives that lead some of the PCs to want to reinforce it themselves. The constraint doesn't have to be a static location per se, though if it involves wandering, the PCs should not be able to simply wander out of the constraint. The constraint or situation should also not be so intense or demanding that it becomes an external goal that unifies the party in confronting it.
Once the constraint is in place, the other critical element is the set of divergent goals. The minimum number is two, but I recommend at least three, so that you avoid simple factionalism and popularity contests. A simple set you can use is to incentivise confronting the situation, escaping from the situation, and a lack of concern or investment in the situation entirely. If you recall what I wrote on a previous occasion about offering goals that only matter to a subset of the party, this is the same thing on a smaller scale. Escaping and a lack of concern work best as goals in this scenario if each goal would require the total effort of the party, or at least the part of it with the least commitment to its success, without benefiting them directly.
A wizard talking everyone else into pressing on further into the dungeon so that he can recover the lost spellbook of an arch-wizard, for which he requires the thief character's assistance to deal with the traps along the way would be an example of this. Meanwhile, the thief can't bail on the dungeon without the wizard's help, because he needs the wizard to cast a spell to unlock the way back out. The fighter, meanwhile, may not care about the spellbook per se, but about helping the thief and the wizard, his friends, whatever they decide.
One thing to bear in mind when designing these kinds of goals is the sunk cost fallacy. If the thief is already in the dungeon, he's already committed at least somewhat to getting the damn spellbook, and the cost of the effort he's put into this already is likely to push him to "complete the quest". Therefore, incentives to flight or disinvestment should normally be stronger in the context of a bottle session than the would be otherwise. However, they too should not provide a clear and present danger. Nebulous, ill-defined threats work best here.
The point of all of this is to get people arguing with one another. There should come a point at which it's clear that the argument is either about to resolve or about to start going round in circles. If it resolves, then great, the PC's collective effort should carry the day, with a reward for their efforts, until the next adventure or at least the next session. If they start going around in circles, then an accidental or unexpected change in the situation should happen and ease the way to one of the goals. You might be surprised at how little justification these sorts of things actually require, as television, movies and novels have acquainted people with this trope such that most find it "natural." Someone leans on a wall and activates a secret door, an external personality finally enters the picture to help them or unify them in opposition to him, the valuable item everyone was arguing over the ownership of is broken in the struggle, etc. The PCs choose how they wish to react to this, and then go on their way, situation resolved.