Thanks for the report Denny. Looking at https://hg.python.org/cpython/file/30099abdb3a4/Modules/datetimemodule.c#l3929, there's a problematic caching of the "_strptime" module that is almost certainly the cause of the problem - it will attempt to call _strptime._strptime from the already finalized interpreter rather than the new one.

It should be possible to adjust that logic to permit a check for _strptime._strptime being set to None, and reimporting _strptime in that case.