Bug report (since resolved) that highlighted our general lack of test coverage for the interactions between environment variable based configuration and command line based configuration: https://bugs.python.org/issue31845

This work revealed the absence, since the refactoring changes the order in which we check environment variables and the command line (we used to check the command line first, and then env vars later during Py_Initialize, now we check env vars in _Py_InitializeCore, and the command line afterwards).

Something I'd also forgotten is that I'd switched the PEP to use "_Py_InitializeRuntime" and "_Py_ConfigureMainInterpreter", but the draft implementation is currently still using "_Py_InitializeCore" and "_Py_InitializeMainInterpreter".

For that last point, it's probably easier to change the PEP back than it is to tinker with the implementation - those specific API names are pretty arbitrary anyway.