When I talk about the build-try-fail loop, I’m talking about the time it takes a player to set up a strategy, try it, and determine it’s success.  The build-try-fail loop applies to any game, emergent or not.  For example, Clash of Clans has a very long build phase (it can take several hours to collect the resources and create the units for a single raid), a short try phase (at most three minutes), and failure is moderately difficult to assess (you can’t build units specific to the defensive target’s setup).  By contrast, Candy Crush Saga has a modest build phase (the time it takes for energy to accrue), a lengthy try phase (several minutes), and failure — which isn’t known until the end — is hard to evaluate due to the random initial state of the puzzles.

The examples above have artificially long portions of this loop for monetization purposes, and it’s hard to argue with their success at doing so.  Nevertheless, the more emergent a game gets, the more it benefits from shortening this cycle.  Specifically:

• Experimentation is encouraged. The longer it takes to set up and make an attempt — a new level, puzzle, raid, whatever — the less risk a player is willing to take.   Reducing this time investment gives players the freedom to try new strategies and explore the entire possibility space the game has to offer.
• Learning is accelerated. The more frequent the attempts, the faster the player will come to understand the game and it’s potential.  And if you agree with the thesis that fun is primarily about learning (solving problems, figuring out new strategies, etc.), then more learning means more enjoyment for the player.  An emergent game with a large possibility space has more available to learn and therefore more potential fun to be had.
• Failure itself becomes fun instead of frustrating. The less invested a player is each time they try something new, the more likely the results will be treated as interesting or entertaining instead of unpleasant.