Which CL "interactive development" feature can't you get with either a jupyter notebook or ipython + autoreload enables? With a smart editor like VSCode or Spyder you also get visual variable inspections, history of executed code in one way or another etc. Only that can't easily save an "image" auto-magically, but data science and ML people wouldn't want to do this other than explicitly (eg. using fast loading formats like feather to reload semi-processed vars etc.).
I don't think you are aware how much the Python workflows of AI / ML / datasci people resemble you "common lisp ideal" ...but with a different language and better libraries :)
And if Julia succeeded, it will also bring decent performance to such workflows.
Sure, the way backend devs or devops people use Python is very static and different, but any experienced ML practitioners are at home with "interactive development".
for the second point: https://mikelevins.github.io/posts/2020-12-18-repl-driven/
There might be more CL libraries than you think: https://github.com/CodyReichert/awesome-cl but I'm not that a zealot to say you can ignore Python's ecosystem, especially in certain areas…