A few months ago, I tried to build in the processor microcode update in the kernel. (Look, it's my kernel; I can if I want.) The code was there, and had been for many versions back, but I couldn't get it to work.
Then, in a submission to the kernel mailing list, a developer posted some very substantial patches, which should, apparently, make the functionality operate as expected. They should be in 4.6 -- I haven't yet checked the release notes.
The thing is, this was code in many "stable" kernel releases prior which can't ever have been tested: it was just broken. It's made me realise that I've had a somewhat rosy view of kernel quality control. After all, there are about 40,000 source files. Linus checks every change.