Absolutely correct. Performance took a big hit in python 3 as well.

If users are performance bound through use-case, chances are they are using pypy. If pypy3 is going to lag python3.6 + why upgrade at all. Pypy just got numpy and pandas.

My sense is that Python 3 was redesigned with web-services in mind like Youtube, when the highest code contributing use-case in python was science and data pipeline related. The science users are all still using python, but only in ways that Python 3 enhancements were not especially important for.