There's nothing quite as helpful as a post telling you that you've been doing it wrong for all these years!And that is exactly where skill challenges fail.
The goal should be to solve the problem you are facing no matter how many rolls that takes. That allows for a lot more flexibility and verisimilitude than skill challenges.
<snip>
As I said: "Roll the highest skill you can get away with".
Skill challenges detract from the actual in game situation by making not the solution of whatever problem the PCs face the goal, but to rack up enough successful skill rolls so that the problem goes away.
Derren, there are two ways to show you're wrong: one theoretical, one empirical.
Here's the theoretical refutation: a skill challenge has the same basic structure as D&D combat. D&D combat does not invovle "solving the problem no matter how many rolls it takes". Players don't describe the ir PCs' brilliant sword play to the GM, who then decides - in light of the fictional situation, like how skilled the enemy is - whether or not the PCs win. Rather, players roll attacks according to a mechanical structure (turn structure, action economy) until the enemy's hit points are all ablated. But plenty of people enjoy D&D combat, and find it more engaging than "rocket tag" combat. For those who do enjoy it, it's engaging because it obliges the players to engage the scene, while preventing everything turning on a single choice or roll. A skill challenge has the same virtues.
Here's the empirical refutation: actually read the play reports I posted, and then tell me where there is a lack of flexibility and a lack of verisimilitude. Until you actually engage with the reality of people's play experience with skill challenges, why should I take seriously anything you say about them?