We know googlebot executes JS. But this could be primarily for things like validating that a site isn't cheating by dynamically hiding search keywords and so on. It is also for generating the page preview.
It's good to see that they're starting to index JS-rendered content too, as seen with the Facebook comments widget, but it does not mean we're free to ignore these issues just yet. As it stands, client-side rendering in general means a huge hit to search ranking and experience (e.g. your search results won't have anything meaningful to say, and will probably use irrelevant text like your static footer/copyright notice as the description).
On that note, I am personally of the belief that the fragments are part of Google's learning/training process for their spiders.
If they sniff the XHR traffic on every domain they encounter a HashBang they can learn lots about the use of AJAX and the types of content being exposed via AJAX.