In D&D 4E combat, the PC:s fight whatever, and either win solidly, win with relatively heavy losses (of HP, daily powers etc), or lose (which means game over, or at least "you wake up in a cell".)
In a D&D 4E skill challenge, you either win or lose. No intermediate results, except for those special "wilderness travel" challenges where someone has to make an Endurance check to avoid being exhausted. The developers seemed to want to do more, but they didn't go further with the system in the first DMG.
A system I've been thinking of using is to divide up a skill challenge in "rounds" instead. Every round, each PC does something (makes a skill check). Once all PCs have acted, the "monsters" act. What this means is that each round that ends without the PCs having succeeded, something nasty happens. This could be a fight (drains resources), or something else. Failures aren't counted, the PCs are just trying to get the requisite number of successes in as few rounds as possible.
Lacking the patience to do the statistics for this by hand, I used Scott Gray's Dice Pool Calculator. Knowing that a trained or talented character has about 80% chance of success, I just have to look for the cumulative probability of getting the requisite number of successes. At first I assumed a goal of 5 successes, because that leaves a slight chance that 5 PCs could succeed before the end of the first round. If the party is smaller or larger, the requisite number of successes should be changed to mirror the party.
As the table on the right (click for bigger picture) shows, it's actually hard to get all successes in the first round, but the group is all but guaranteed to get them in round 2. Larger groups obviously have a harder time if everyone must succeed. If you really want to prolong the challenge, require twice the number of players in successes - they have a small chance of winning in round 2, and are likely to win in round 3.
Weirdness with large groups aside, this does do what I'm after. The PCs are likely to see what happens after a round they've failed, but they have a chance to avert it, and it probably won't happen twice. It seems to work, especially for "investigation"-type challenges, where the characters do something for a day (make a skill check each) and then the bad guys send assassins each night (Fight!).