What if the infringing content was committed after the fork was created? It'd be pretty bad locking repos that did not merge in the offending line(s) of code.
That's a sensible implementation if you want to be sure you're covered (or aren't ballsy enough, I'd probably take out the forks), but I wonder how long they could go by not taking action on the forks unless the specific forks are identified.
Just consider Youtube a moment. Couldn't they minimally run a hash on any uploaded/embedded sound and see if it matches the hash of some taken-down material before? It'd probably catch at least a significant number of users even if it's trivial to bypass and susceptible to compression losses and quiet-extensions and so forth. (An AI could always "listen" to it though and probably perform pretty well.) Anyway, Youtube doesn't make an automated effort to take down infringing material (that I know of) so I don't see why Github has to.
Assuming a modicum of technical sophistication on behalf of the legal people (which may be an invalid assumption), I think the fact that git/github is centered on the forking/sharing of code with an explicit graph to walk would just incur too much liability for them because it is obvious (in a technical sense) where the code has gone.
You don't have to do any forensics in the sense of your youtube example, you just have to follow the graph.
Whether that it is a legally compelling point, I have no idea.
I imagine they have to lock / disable all forks beyond what the notice mentions as well, unless Sony, et al don't understand how git/github works.