Something that’s been under discussion for a while is splitting off the scalar kernels for ufuncs and gufuncs in scipy.special into a separate library. I’ve created a new repo, GitHub - scipy/xsf: Special function implementations, we could use for this purpose, and have written up an RFC with motivation, a migration plan, and a rough roadmap for the rest of the year here, RFC: Plan for establishment of xsf library for special function scalar kernels · Issue #1 · scipy/xsf · GitHub.
I’ve also created a SciPy issue here, RFC: Splitting off special function scalar kernels into separate library · Issue #21649 · scipy/scipy · GitHub. My intention is for the SciPy issue to be a place to discuss whether it’s a good idea to split out the scalar kernels in general, and what this may mean for SciPy, while scipy/xsf#1 is for discussing specifics of the migration plan and roadmap. This isn’t meant to be strict though; feel free to comment here or on either.