A few years ago I really got into Indy RPGs, especially the game design theory behind them. I hung out on the Forge forums, read "Design patterns in RPGs," all that jazz. And one of the game design concepts I took away from that is the difference between task resolution and goal resolution. The short version is this: task resolution is a system for determining if characters accomplish what they set out to do, while goal resolution determines if what the characters do gets them what they want. Most games, including FATE, really only deal in task resolution. They leave goal resolution up to the narrative, usually controlled by the GM. But I always have a soft spot in my heart for goal resolution mechanics.
In FATE, you have three different ways of handling task resolution: Challenges, Contests, and Conflicts. Interestingly, though, the challenge is the basic building block that the other two types are made of. You could say that a contest is really a number of back to back challenges, or that a conflict is made up of various challenges with many varied results. Lets take that line of thought further. If a contest is made up of a series of challenges, could it not be made of a series of more complicated tasks?
This is the Contest Fractal: each exchange in the contest can be resolved with either a challenge, a contest, or a conflict, depending on it's nature. It's a way of using a contest to frame not just a scene, but a whole story.
Start your story as you will, at some point your players will have a goal. Get them to clearly articulate that goal, like "we will save the princess." Excellent. Now figure out who opposes that goal, such as the evil duke who kidnapped the princess. They've got their own win condition, say marriage to the princess. Now that the opposition and goals are clearly stated, we can use the contest rules to give our story a framework. Each exchange in this contest is going to be a scene. Each scene will have a winner and a loser, but the nature of the scene will determine if that requires a challenge, a contest, or a conflict. The first scene might be sneaking into the duke's castle, handled with each player making a simple stealth challenge. If the players succeed in sneaking in, they tally a victory and we move onto searching the castle in the next scene. But let's say they fail. Then our Duke tallies a victory, and the next scene is a conflict with the guards. They players win or lose the conflict, tally the appropriate victories, and move to the next scene. Let's say they lose the fight against the guards, too. Duke has 2 victories, and the players are captured in the dungeon. The next scene is an escape from prison scene. If the PC's fail this one too, they're in trouble: the Duke will have tallied three victories and the players were too late to stop the wedding! However, if they succeed, they'll tally a victory, and hopefully succeed in the next two scenes as well.
You can finesse the framing here by using alternate contest rules, such as moving the goalpost, tug or war, or cat and mouse from the DFRPG. In this way, the GM can fine-tune the contest to produce the kind of story you want. Again, the basic idea here is that each exchange in the larger contest is one scene, with success in that scene resolved with a challenge, contest, or conflict of it's own.
What this gets us is a solid mechanic for determining when players achieve their goals, or when those goals become unobtainable. It also gives us a decent idea for when milestones should occur. Win or lose, the end of a story contest is probably a milestone. For episodic storytelling, it tells you when the episode is over too. This won't be everyone's bag of tea, of course. Many storytellers would hate being railroaded by these kinds of rules. But for others, having this kind of plot-framing guidelines makes running things so much easier.