> and perhaps Python-using academics also never make packages.

Your guess is (largely) correct. For most users, the workflow is exactly as you described:

> They make the notebook, use the build-in R functions and then their work is done.