Here's where I think I should point out the nature of the DM knowing their players. In general, I tend to do four things:
I base DCs on what I know my players are capable of achieving. If the PCs have low levels of say physical skills, I either make it clear that I expect them to somehow be able to achieve relatively difficult DCs or I set the DCs based on those skills when it is obvious a skill challenge requires it.
I don't think this is 'wrong', far from it in fact, but its an area that deserves a little more consideration. If you constantly just cater to what the players and their PCs are good at and comfortable with, you're either creating a kind of 'cake walk' (the DCs may all be at the recommended levels, and the encounters perfectly difficult, IN THEORY, but in practice easy your group). The other possibility is that you create a self-reinforcing optimization spiral where the players see only really value in investing in what they use most, and the DM responds by providing more, and more difficult, challenges in just that area.
The alternative is to provide a wide spectrum of challenges. Let the players decide which ones they want to engage with, and which they don't. Let them decide which types to optimize for, and which not to. Just keep giving them all these different choices. This reduces optimization and can keep the game more fresh, but it could also just make players board or disinterested when they're asked to do something that they don't care about.
Its a different way of saying to know your players, in the end. The thing is, I don't think 4e is in any sense unique here. SCs just present situations. You'd have had to know these things about your PCs just as much back in say 2e. In fact I think 1e's OA was probably the ultimate place where this was true, as many skills were 'trained only' and you really didn't get much chance to add more training. Plus a lot of them were kinda weird stuff like 'flower arranging' that had specific value in the context of the milieu, but weren't inherently all that interesting in a game sense.
I tend to throw in really easy DCs or even auto-successes into my skill challenges - just because a party is high level, they can still find a 10' cliff to climb or an easy lock to pick - it just isn't the focus of the challenge. It helps the PCs understand how far they have come and also reduces the sense of treadmill scaling challenges. A routine poison needle trap that almost killed the Rogue at low level that makes a return in Paragon allowing the
Maybe those things should just be 'ordinary activities' (IE you don't make checks for even level 1 PCs running down corridors, level 20 PCs probably should just swarm up 30' natural cliffs without even needing a check). Not that it is a big deal either way, really.
I try to make sure that when I design a skill challenge for my group that each PC has something interesting to do. And if the skill challenge doesn't provide that, then I'll redesign it. That's a huge problem with skill challenges in mods as they basically tend to assume that a party with all basic skills is there.
Its certainly nice when this is the case. OTOH SCs TEND to go reasonably quickly (faster than combat) so I'm not so worried about spotlighting a PC in one, though I certainly try to make the more in-depth SCs have parts for all, as you say. Truthfully though, with 17 skills, and say 7 of them active in any given SC, it should be sufficient for an SC designer to have 1-2 physical, 1-2 knowledge, and 1-2 activity skills present in each challenge, as that tends to pretty well guarantee everyone has something they can try at least a couple times. Also group checks (easy DC usually), AA, and possibly additional alt skill use providing things like 'unlocking' and such should allow filling in where the spread is weak.
Finally, I think about how the party might fail forward. So as an example, to use the rapids as an example. If the party isn't doing well, maybe a combo of perception+dungeoneering sees a tunnel to crash into rather than be dragged underwater. And that tunnel might have some interesting bits of information as a consequence of 'failing' to get to the destination on time. Or they get dragged underwater and miss out on that bit of info, but instead manage to save themselves by finding some ancient boat wreck, with a skeleton who has a sigil ring on its hand. Which when they get to the city, they can find out belonged to some minor noble house. Which could then lead to the knowledge that the heir was murdered by a usurper many years ago and get the PCs involved in some political intrigue. Etc...
Yeah, there's many possibilities there. Honestly though, I'm really in the school that says "its no different from combat, let the PCs put it all on the line" and not worry TOO much about the forward part. I mean the trick is really to avoid strict bottlenecks, and this goes along with letting the players choose. They choose the river, or they could try to approach the hermit druid and see if he can help them. Or invent a ritual to call down the giant eagles, or etc.