Here's my wild prediction: 10 years from now, with the benefit of hindsight, we will trace the moment where Google's Page Rank empire started to crumble to the introduction of this html5 feature.
However, discovering this makes me wonder what other incredibly useful but as yet unknown html5 features are out there in the wild?
I don't know how much Javascript the GoogleBot currently parses and/or executes, but I imagine it would be enormously complex to embed v8 into the crawler, and then execute all Javascript on every page and determine what to index on a "page." If we ever get to the point where you can't screen scrape a web app without also executing its Javascript, then PageRank-like methods become less effective, and we'll need more semantic approaches to search. It's a great thing such semantic approaches don't yet exist!
Also, Google is already promoting a convention for making the various parts of an AJAX application reachable via different URL-#fragments more easily crawlable. See:
http://www.google.com/support/webmasters/bin/answer.py?hl=en...
A number of sites including Facebook have adopted this convention; app developers usually want to be found by search engines.
It's funny you mention Facebook. I would say Facebook's chief motive right now is to stop Google from encroaching upon their territory, so Facebook wants to put up as many speed-bumps as possible around Google. Basically that means keeping Google sandboxed—it's okay to let Google index anything they want and throw that up on an ordered SERP, but no way can we let them grow elaborate ivy APIs and web-based UIs around our walled-garden of user generated content. (i.e. Buzz.)
Given Microsoft's ownership and early partnering with Facebook, Microsoft and Facebook have aligned interests in keeping Google in the sandbox.