Raymond Hettinger added the comment:
> Some parts of the patch are harmless but other parts needlessly constipates the code for very little benefit.

In general, Python startup is currently the most severe performance
regression in Python 3 compared to Python 2.7:
https://speed.python.org/comparison/?exe=12%2BL%2Bmaster%2C12%2BL%2B3.5%2C12%2BL%2B3.6%2C12%2BL%2B2.7&ben=649&env=1&hor=true&bas=12%2BL%2B2.7&chart=normal+bars

Python 3.7 is the fastest of Python 3 versions, but Python 3.7 still
more than 2.2x slower than Python 2.7. 2.2x is a large number, it's
not a tiny slowdown.

I agree with Serhiy that argparse became more popular so became more
important for Python startup time in applications.

Maybe Serhiy can try to run a benchmark for get timing (milliseconds)
instead of comparing the number of imports?

Sorry, I didn't have time to review the change itself yet. But I like
efforts to optimize Python startup time in general.