The original motivation from @rth: "One point is that this needs numpydoc master, and generally I would rather see it as an optional linter than unit tests. Currently this validation is run in one of Azure CI jobs, maybe there is a better way to do it. The code style in the tests can also probably be improved."

I don't have a strong opinion but since we tend to include our consistency checks in the test suite I don't mind adding this one as well.

The original motivation from @rth: "One point is that this needs numpydoc master, and generally I would rather see it as an optional linter than unit tests. Currently this validation is run in one of Azure CI jobs, maybe there is a better way to do it. The code style in the tests can also probably be improved."

It's now using a released version of numpydoc. So at least that specific concern has been raised.

For the other points, I think having too much complexity in our CI setup is detrimental, both for maintenance and for understanding by users reading the report.

I checked in the Azure test report view that the tests are being executed as expected by the pylatest_pip_openblas_pandas config (they are skipped otherwise because numpydoc is not installed).