I wonder if you could automatically a/b test the impact of changing the og:images tag has on social media sharing/interaction. That would be a nifty feature too.
I think you cannot test on the basis of one article, but in general, you could swap between entirely different templates - hmm I should write that down