Does nbdev paradigm clash with only-one-jlab install?

To reduce install footprint/time/env-complexity I would like to switch to only one jupyterlab install per conda installation, so, put JLab and nb_conda_kernels into the base environment, and all envs only get an ipykernel package, so that the base jlab can activate all other envs kernel.

Q1: Will this work with nbdev or does it need a full jupyter install in each conda env?
Specifically for checking functionality of the nbdev package I develop in other python minor versions (3.9, 3.10, etc.), I would have one conda env for each to see if things are running.

Q2: Are there better ways to do this? :wink:

unfortunately, dask also clashes with the one jlab in base install paradigm, so i might just as well give up on that. :frowning: