It's just the servers. The kernel repositories weren't modified, and even if they were, git makes it pretty impossible to modify history if you're paying attention -- the next push would fail as a non-fast-forward.
Also, in the mailing list thread, someone actually went back and verified the contents of all the old tarballs and compared them to the bits from Git, since otherwise someone could have changed the tarball releases.
And the HN story referencing that announcement: http://news.ycombinator.com/item?id=2947578
I have to wonder why someone felt the need to anonymously pastebin the original announcement to a private mailing list.