Thanks! cc @ycopin

Needs a change log.

Does this need a test as well? Why wasn't this bug caught the last time?

It's not technically a bug because the code does what it's designed to do, it's just that design is wrong. So a test won't really help (IMHO) because if there had been a test that the change would cause to fail, the claim would have been made that the test was wrong because it was testing the wrong thing. In the original issue, there was a request for a proof of the revised (incorrect) formula, which was never presented so I guess there was a bit of misplaced trust.

I've put the PR# in the changelog (in the right place at the second time of asking) and squashed commits so this ought to be good to go if you're happy with the algebraic proof presented in #8094

Thanks @chris-simpson! I think the confusion came from 1/gamma**2 being included in rr_gg.