4e is supposed to be about everyone getting to play and opt in and such. Even if we assume that class skills are well balanaced (they're not), there's still too much of a case where people can be left out, or end up only ever using their one, best skill.
I'm opposed to the current system which says, 'if you're not sure you should be rolling, don't roll, do nothing- or you might earn your party a failure point'. I'd much prefer a systme which explicitly says, "your turn is an asset and even if you aren't doing that well in this context, you can be sure that by using your turn, you are contributing'.
Honestly, I just feel liek skill challenges are a missed opportunity, and should have been a lot more colaberative and improvisational.
I would just say run them that way then. I do. I mean it is going to depend highly on the group how exactly you work things anyway. I can say to the people I play with "Hey, what do you think should be in here?" and the answers will be interesting.
Most skill challenges also need to really affect some kind of change in the plot that is at least somewhat meaningful. The N success before 3 failures is also just a very simple basic default mechanism. Both DMGs make infinitely clear that the concept is you can add any different mechanism in there you feel like you need. Just like you can invent terrain, monsters, etc for a combat encounter you can pick some specific mechanics for any given challenge, make up your own, use some 3rd party thing, etc.
You can use Obsidian if you wish, where everyone always rolls and the whole thing is like a group check, where everyone rolls 3 times. After the 3rd roll the total successes are checked against a 'DC' based on party size. I have found it is good for certain things, but not useful for others. The basic system does some things perfectly well, and there are other easy to figure out possibilities.
You can have a graph, which could be good if you cannot justify any kind of tally, but the issue with it is you've devolved things down to 1 toss of the dice again. This is why checks are normally pooled to whatever degree. Still, it is possible, though really better as a meta-challenge.
You can have neverending SCs too, like the way I would do Mike's Suderholm example in DMG2. Just drop checks off the end of the list, so when 3 in say any 7 checks are fails then the guard shows up to arrest you, etc.
Framing is important. You want to incorporate a hunk of action that provides some variety of subtasks that use different skills, but which makes sense to apply a pass/fail to. The 'timer' SC is always your basic fallback, or one that just steals surges (which I think is dull). Can we open the lock in time to intercept the Doppleganger? lol.
Negotiations make good SCs, there's negotiating, spying, dealing with the other guy's spies, being spies, etc. "You must get X piece of information from the enemy without them knowing you have it." Meanwhile also negotiating.
A book on the subject would be good, it would seem, but I really think it is kind of hard to explain, perhaps especially to a lot of people that have basically played D&D for the last 20 years, lol.
Don't be afraid to do things besides SCs anyway. You can make up quests that will surely require a certain number and types of skill checks to succeed on the way, with an XP reward equal to an encounter/challenge. Same thing just disguised as something else.
