Is that farfetched? Does Google not have the engineering resources to detect the image canonical source (a la Tin Eye) and point to that instead?

I find that hard to believe considering the resources they’ve put into Content ID at YouTube for copyright infringement identification and takedowns.