To me, the most surprising in talking to sources in Japan was what appears to be a big shift in the local industry's thinking. Details of their new ITS roadmap are expected to be discussed next month at ITS world Congress.
Seems to me that to support true self-driving cars, you will need both types of comms. So if budgetary constraints slow the introduction of vehicle to infrastructure, they'll also be slowing down the introduction of self driving cars.
As to natural disasters, they do tend to damage fixed infrastructure in a way that is more difficult to recover from. But I wouldn't conclude from this that therefore vehicle infrastructure communications aren't very important. For a self driving car, they're still indispensable.
I suppose you can concoct some weird and chaotic schemes, like all moving cars creating this gigantic ad-hoc network, where data from each car's local radar sensors become aggregated into a huge database of road conditions. Or the other way around, where the local infrastructure always intrudes into V2V comms, so there's never a need for that direct link. Aside from the greater complexity, it's not hard to show how either approach has its limits, if you expect self driving to become reality.