This caught me at first too. My guess is that they are blocking access to the infringing file for people other than the alleged offender.
Previously they disabled the whole repo, and this blocked everyone, including the alleged offender from accessing it.
Now it sounds like they will restrict access on a file level and give the alleged offender access to attempt to respond/modify the file to comply or appeal.
This effectively removes the offending content from public view, which should be interpreted as blocking access.
I'm not sure if this is exactly what happens but from how the process was described, it seems plausible and avoids some of the worries that Github isn't responding fast enough to claims.
Is there a mechanism they could use for blocking access to the file from within git without rewriting all of the commits (and therefore changing hashes)?
I'm not sure I completely understand your question, but my guess is they are blocking access to the file from github.com which is all they need to do to comply with the DMCA request. They can't do anything about people's local copies.
Sure, they can block access to the file from their web interface. I'm asking how they block access to a specific file when someone does a "git clone" or "git pull".
It sounds like they just wait 24 hours to get contact.
"because GitHub cannot disable access to specific files within a repository, we will contact the user who created the repository and give them approximately 24 hours to delete or modify the content specified in the notice. We'll notify the copyright owner if and when we give the user a chance to make changes."
Previously they disabled the whole repo, and this blocked everyone, including the alleged offender from accessing it.
Now it sounds like they will restrict access on a file level and give the alleged offender access to attempt to respond/modify the file to comply or appeal.
This effectively removes the offending content from public view, which should be interpreted as blocking access.
I'm not sure if this is exactly what happens but from how the process was described, it seems plausible and avoids some of the worries that Github isn't responding fast enough to claims.