As far as I know almost always coverage is used to see "test coverage".
For unused branches/variables etc there are linters, dead-code finders (such as vulture or dead).

So I don't think code coverage is important to extent of running run_all.sh script.

I would like to

• remove run_all.sh
• collect coverage
  • only for tests
  • only for one python branch (e.g. 3.7) (although it is possible to do coverage run -p and then coverage combine)