First, Python2 need not be EOL. I’m clearly talking about some parallel timeline where python as a *nix shell became a project and got traction. Somewhere around ~2000, when python2 was first released, some project works on making the necessary adaptations for a shell and presumably gets that working well.
Second, we’re talking about Bash and Bash scripts, any serious dependency management or package management doesn’t really factor into this, it seems you’re bashing Python as a fully fledged programming language.
The weakest point in my imaginary proposal, IMO, is getting python to work as a shell language. As a scripting language it’s fine and many are happy using it where Bash might be used. But as a shell, and in scripting too, it’s interop with core utils and commands is pretty suboptimal. Ergonomics would at some point require something to be done about that and I don’t know what a successful version of that would look like. If we’re talking about a POSIX standard, maybe all that’d be necessary would be a standard library that wraps everything needed? Or a nicer object/function/syntax for calling subprocesses, which various third party libs have tried obviously. The Xonsh project also comes to mind … a superset of python, which is a great idea IMO but maybe undermines the idea that python2 is everywhere already.