Thanks for the feedback.
Yeah, so I am imagining having dagit running serving a bunch of different users and projects, so ideally we would want to be able to separately develop user pipelines with minimal touch points. So that one users broken pipeline doesn't break everyone elses experience.
I can see that this could be organised by having a single repo that references a bunch of namespaced pipelines. So that the core pipeline code is all separate, and the only touch-point is then the central repo.