Looks like that's exactly what Meteor's spiderable package does since 08/2012[0]: look at user-agent, run phantomjs for 10s and return a rendered page once google/facebook crawler detected.

[0]: http://www.meteor.com/blog/2012/08/08/search-engine-optimiza...