I passed my scaffold question on to my 20.20 professors, and got a very thoughtful response. (To their credit, it was also a very timely response; I've just been slow in posting it.)
Recall my question: there seemed to be a conflict between the role of kinase cascades as signal amplifiers, and the role of scaffolds as giving specificity. Turns out, naturally, that there isn't really a conflict at all. Rather, I didn't fully understand what scaffolds do. The crux of what I wasn't understanding is that proteins go on and off their scaffold all the time -- they don't stay bound (semi)permanently.
One of the main benefits of scaffolds is to make signaling pathways more efficient: they help bring proteins close together so they can interact. If you have only a few molecules of Protein A and Protein B, then they may not encounter each other very often if they're just freely diffusing around the cell. A scaffold can grab one Protein A and one Protein B, hold them close together long enough to interact, and the release them and go look for two new proteins. Or if A is a kinase for B, then A can stay on the scaffold and multiple copies of B can take turns.
Scaffolds also act like little switchboards, dictating where signaling pathways are allowed to cross and merge or where they must stay separate. Suppose a scaffold has one binding site for the kinase A, and a second binding site that can fit either Protein B or C, but not D. In this case you get activation of both B and C, and the signal from A propagates in two different directions, provoking two separate responses (but not provoking a third response). This is really helpful because a lot of kinases are promiscuous -- they'll phosphorylate anything they can get their hands on -- so if they're confined to scaffolds, then they'll do whatever the scaffold says they can do, and no more.