I am referring to the adaptive block size limit proposals floating around lately with bitcoin. That's the OP's question about scalability, as in transaction throughput scalability. What I am more worried about is block chain size scalability, as in initial block chain download, as I believe that would hamper bitcoin's ability to scale (at least in the peer-to-peer sense).
Yes - it's a connected issue. You have
a) block updates
b) chain bootstrap
a) is the differential of b). For the "block chain size" to be scalable, any client must be able to bootstrap in a reasonable amount of time - ideally through an internet connection, i.e. the bitcoin network, since any private boot-straping exposes you to counter-party risk. So I agree with you on this issue. However, there are slow and fast internet connections. Thus one may be able to bootstrap the node on a fast connection, while maintaining the sync through a slow connection. However, it is crucial that clients are able to keep in sync with the network even with moderate internet connections (sub 1MBits).
So I'd say that b) is a weaker condition than a).
Maybe you were talking about block chain size as well, I am not sure. Or maybe some design that doesn't store the full ledger at every node. I don't know, that would be quite a challenge to do.
Yes. In fact I was talking about strategies to change the scaling behavior of the consensus model - but scoped to the client. So the whole network may still be allowed to scale with O(Users*TransactionsPerUser). But individual clients may not need to know the full state of the system. It only has to be good enough to prevent double spends - no solutions seen yet.