I agree that there is (at least) a certain delicacy in reconciling 4e's p 42 and its powers. But that is (in my view, at least) mostly a consequence of 4e being very specific with its powers. Games like HeroWars/Quest, or Marvel Heroic, show how you can have systems which are at one-and-the same time exhaustive while leaving plenty of room for out-of-the-box situations.
I'm not entirely sure "exhaustive" is what I'd call Marvel Heroic. I've read through the rules as I was super excited to see how it did things after hearing so many good things about it. I've watched youtube videos of people playing it to see if I was missing something, but in the end all I can figure out about the game is that it's a dice adding game with a story attached.
Being Wolverine doesn't actually give you any benefits at all. Your claws don't do extra damage to enemies compared to people without them. Your regeneration doesn't have any real effect over someone's forcefield. All of them are just dice that get added together in a pool and your individual powers are inconsequential most of the time. If two dice pools get added up to the same amount, the powers that make up those pools don't effect the game more than the DM chooses to make them effect the game. Which is normally a short description about how you hit the enemy with claws or an energy beam.
The mechanics are so disconnected from the story they create that they don't seem to have any relationship at all. It's possible to have a mechanic that's completely exhaustive and handle out of the box situations when the mechanic is "Roll a d6, on a 4-6 you win". All you need is a DM to tell you what "you win" means in any given circumstance.
However, if your goal is to have mechanics that are more tied directly to the action you are attempting, then the more tied you are directly to what is happening the less they cover corner situations. If you say "All attacks follow this formula" then if you need to now decide whether each action taken is "an attack" so you can decide whether to use that formula or not. If an action is somewhere between an attack and something else then it becomes hard to adjudicate. Like the trip example. If you are just attempting to trip someone do you make an attack roll to do so? Or if you make a strength roll high enough to literally "pull the rug out from under them" do they just fall over automatically? What's the DC to pull the rug? I know my ruling would be something akin to "you can't trip him that way unless you have a power that trips them" so that I don't have to constantly come up with rules and I avoid making other players feel like their choices weren't worthwhile.
It's a bit of a tangent, but I think the problem with this is that the GM isn't doing his/her job. S/he isn't introducing complications, or developing the situation, in a way to which the players can respond. If the GM has nothing interesting to do with his/her NPC on a failed Diplomacy check, s/he shouldn't be running a skill challenge! (Robin Laws emphasises this point in HeroQuest revised, and that bit of the rulebook was cut-and-pasted into DMG2.)
The problem isn't that there's nothing interesting planned on a failed Diplomacy check. It's that a skill challenge has a distinct structure. You must succeed in, say, 12 successful checks before you "succeed". Often, it is hard to find an excuse within the game to even make that many checks.
In practicality, skill challenges just couldn't be done without it feeling "out of place" beyond a certain complexity level. You couldn't have a skill challenge to "convince the king" of something that required 12 successes because play degraded into continually rolling diplomacy checks. None of the other skills had a very logical reason to be used in this circumstance and any attempt to use other skills always seemed like a leap of logic and that the player was trying too hard.
Most of the time we attempted skill challenges they always turned into a guessing game: "I'd like to try...umm...religion to convince the king." "Sorry, he isn't that religious and won't be swayed by religious arguments" "Ok then...goes down the list...I attempt Perception to point out how beautiful something in the room is in an attempt to flatter him."
The best Skill challenge I ever saw used was SO disguised that no one knew it was a skill challenge. It took nearly an hour to run and involved going to 6 different locations, interacting with a bunch of different NPCs and a number of different situations all to "discover where the enemy's hideout was". But because it was so disguised, you has to reach a certain point in the narrative in order to succeed at the skill challenge anyways. Which meant you often had to get 10 or 12 successes before you got to the person who actually had the information though the skill challenge itself only needed 8 to succeed.
It felt like an unnecessary construct that only served to confuse the issue. If it makes logical sense within the narrative that you've found the hideout in 4 successes, let it take 4 instead of the 8 you had planned. But if you do that, then might as well just throw out the skill challenge as a concept and simply ask for skill checks when they are appropriate.