Nice double-post...
1) How would they pay for network fees (= who gets the forkXRPs)?
2) How would they ensure a healthy validator network structure?
These are the 2 main remaining questions for mainline Ripple too...
Is running a Ripple-esque network not feasable if decentralization was the aim?
What about the new Stellar consensus protocol the SDF is working on?
Sure it is, they can just run mainline Ripple as easily though.
It only renames a few components (UNL = something something slice) and seems to be a nearly 1:1 copy of Ripple Consensus. It is nice that they expose the UNLs to the network but apparently these are not checked or used to verify trustworthyness anyways by other nodes, so all that is left is a more "bucket based" approach than with Ripple, which rather wants to approve sets of transactions in fixed maximum intervals instead of moving forward in all undisputed areas while waiting in disputed markets.
This would have no major impact if they were using Ripple (or a fork of it).
I don't see how the lack of an independent fork is the thing that keeps exchanges from using Ripple as is... mainline Ripple however might soon have more banks, Western Union etc. on its ledger, so an inter-exchange only Ripple fork might not be very popular to begin with (and not very resilient to legal pressure, depending on how it is run).
Probably because Ripple Labs controls most of the XRP. Ripple is great and all, that's why there could be more benefit for the exchanges themsleves to fork it and build their own Ripple-esque network and divide the fork-XRP among themselves to give away to their users.
edit: Imagine if (just if) Bitfinex, Bitstamp, BTCe, BTC China and all the others joined together and built their own Ripple-esque network. I think there's a good chance it will be very popular.
I wouldn't want to use a ledger run by Bitcoin exchanges exclusively, but then again who knows... apparently e.g. Bitfinex even uses Tether (which is more or less a gateway built on Counterparty built on Bitcoin). Inter-exchange fast path payments are a years old concept which was first commercialy tried by Charlie Shrem's company Bitinstant... well, it didn't end well to say the least.
Gateways don't need to care about XRP prices or distribution, for the price of a hamburger/coffee you can get enough XRP to run a gateway almost indefinitely from a current perspective.