With a convoluted enough kernel construction one can draw a boundary between any set of points. I think technologists consistently misunderstand the utility of the latest and greatest. Young technologists are especially prone to overvaluing the latest thing even though in the grand scheme of things there is almost no or even negative value in changing whatever currently works but they convince themselves there is real value in remaking the thing that works.
The reasons are arbitrary and are post-hoc rationalizations about the latest technology and when an example is brought up about the IRS still running on assembly and cobol the young technologist creates the convoluted kernel to separate the IRS from their special use case. There is no good reason to upgrade a python 2.7 codebase to a 3.x codebase. If you think there is then you are probably fooling yourself because you are young and foolish.