In this regard I’d agree with @Patches: we don’t intend for anyone to clone the git repository in order to use Certbot as an end-user. The versions of Certbot that exist in the repository are not considered “released”. In rare cases, they might even be missing functionality that we consider important or have known bugs that are worse in some way than the released versions.
Certbot releases (and packaging) usually involve reasoning and discussion about what functionality should be included and whether anything needs to be modified to integrate with a particular OS—and, at a bare minimum, deciding when enough new functionality has been added to merit a new release.
The git repository exists to promote coordination among people working on developing, debugging, and documenting Certbot, including the Certbot team at EFF and outside contributors. It’s optimized for those purposes, not for other purposes.
As @Patches referred to, we have tests that are run directly by the developers as well as by continuous integration systems that are integrated with the GitHub version of the repository. Some of the CI tests are meant to confirm that changes made to one part of Certbot didn’t harm the functionality of other parts. For most purposes, it’s useful to have as much related code as possible in one place so that the tests that get run in response to every change are as comprehensive as possible.
While a CI system could certainly be set up to test things together that are built from different repositories, I don’t think we’ve seen this as a priority and I think it could make developers’ lives more difficult.