The only _real_ examples where there would have been a scaling advantage were centralized networks run by Ryan (and maybe a handful of other servers).  I suppose you could call RipplePay and Villages proofs-of-concept with the possibility for a federation glued together by Bitcoins, but afaict nobody ever implemented anything like that.  It was just a series of vague proposals on the mailing list.  (And even if someone did implement that, they would run into the same centralization problems Diaspora has without any of the data control features.)

If you wanted to create a bonafide p2p Ripple without the use of a blockchain, how would you deal with nodes appearing and disappearing just as quickly as they do with Bittorrent?

I'm not saying that the new Ripple solves this-- I'm saying it's the next logical step in a system that feigns decentralization while it implements a centralized solution.  The current Ripple XRP system and (to a large extent) Freicoin are both _centralized_ approaches to digital currencies.  Whether its solving puzzles, downloading/verifying a blockchain, or maintaining the node list, why should the user do any work whatsoever without knowing the rules for how the rewards (or the bulk of the initial rewards) controlled by a single entity are to be divvied up?  This would be like if Bittorrent had started by encouraging hundreds of thousands of nodes to connect through a bunch of trackers, then at some point in the future having a single node with all the files people want start uploading to everyone else.  It's an absurd idea for bootstrapping a network-- or, more properly, it's a way of implementing a protocol without having addressed one of its core problems.

You can't claim to have designed the next generation of digital currency and punt on how to allocate your resources.  Of course I'm happy to be wrong if it turns out that, "just ask and we'll give you some," is the holy grail here.