@churchyard suggested incorporating Rawhide bugs.

The reason I excluded Rawhide bugs is that it's harder to "bucketize" them, and a lot of my motivation was looking at how things changed over time. The original motivation for this was looking at EOL closure over time and I branched out from there. I could certainly look at some Rawhide stats, I'm just not sure what would be interesting.

I suppose one thing I could do is look for Rawhide bugs closed between the branch dates and include them in the corresponding release section. I don't think we'd want to combine them for all of the analysis, but it could give us a measure of "bugs that never reached the user".

There may be other places where adding Rawhide in makes sense.