The current reported code coverage stats on codecov.io miss some global statements of modules imported before coveragepy begins. This is documented in the dev guide, and the suggested fix is here: https://devguide.python.org/coverage/#coverage-results-for-modules-imported-early-on but that recommended fix doesn't seem to be included in the CI.

Using the recommended 'hack', the coverage report shows an extra 4k lines are coverage in Lib, increasing coverage percentage by around 3%. 

Having a more accurate view of the coverage makes it much easier to see which parts of the code need help with tests, without having to go into every individual report to see if global statements are misreported.

I'll raise a PR for this shortly