Thanks for the review @RonnyPfannschmidt.

Something seems wrong with the Codecov coverage notes on the PR diff, it seems to consider every new line as uncovered even though it definitely is. I may look into it later.

I've considered backporting this change, but decided against it since it might lead to more regressions that I can't think of. Instead I think we can release 8.3 somewhat soon (I do have some more regressions to look into...).