@f5937b5b I had also looked at potentially rebasing on Fedora for the image base as well, so we could use more common tooling and not deal with the occasional interesting Ubuntu quirk.  Basing on Jupyter upstream was ultimately less overhead and allowed us to more easily contribute to upstream repos/projects.  It's not uncommon we find new bugs before the upstream maintainers themselves do with our CI process.  It's nice to find those and pitch in *before* we need to deploy it ourselves. 🙂