There is no significant correlation between programming language and bug rate

This is interesting but there are some above (Perl, C++) and below (Haskell, Scala) expected values based on the number of commits. I also am not sure this methodology is the best way to address the issue (although I'm not sure it's the worst either). If for some reason bug fix commits were underrepresenting actual bugs, you could get some distortion.

It seems reasonable to me this is all accurate, but I'm also not sure it's quite the end of the story.

Also Rust is missing. Not that that invalidates anything but it seems interesting as an important case given that part of its appeal is specifically in minimizing errors.

Perl is almost twice the average.