I think the fact that it's a distributed model is the scariest part! One of the biggest challenges of creating the new apps will be exception code for those distributed services that aren't working.

Perhaps you control every service and can get them all back on line if one goes down, perhaps you don't use a T1 or you have one that NEVER goes down, perhaps encoding, http, parsing and internet bandwith will always be fast enough, perhaps the interfaces to your services will never break...

Distributed can mean slower with more points of failure and less contol over them.