This is kind of a needle-in-a-haystack type problem for us -- there's no easy database query I can do to isolate an individual large commit.
The graph you see is the result of a monthly summary across all 8 Subversion repositories, which means I can't give you much accuracy on where or when the massive checkins come from. I can tell you that the first sharp rise occurs in April 2004. Check out this url for some more clues.
I've done some of these investigations before in the course of looking for bugs in our downloader. They usually turn out to be either the addition or deletion of a large third-party library, or sometimes the project documentation has been added or removed from source control.
Occasionally, it's generated code being added to or removed from the source control.
If it's a single large file there's an oustide shot I can find it in our database, but if it's a commit with a large number of regular-sized files there's not much hope I can find it easily. I don't have much better tools on my end than you do on your end -- I'd start by scrolling through the Subversion log of April 2004 looking for a massive checkin.