Bugs per commit

It looks like the commits that introduce the most bugs are mostly the ones that add new drivers, which makes sense in that they also introduce more new code, in general, and also generally don't have any tests for corner cases the author didn't think of (unlike changes that often have tests that someone wrote previously). And they generally can't cause regressions (other than hardware users didn't need waking up and causing problems), so the bugs take longer to notice.

You could reasonably say that the addition of the xe driver broke down the single big bug (of the hardware being entirely unsupported) into a lot of smaller, harder-to-trigger bugs and fixed a lot of test cases while leaving a relatively small number of test cases unaddressed.