I think it really depends and it's a case where a lot of engineering judgment and taste comes to bear. For example right now I'm maintaining a Jenkins system th

I think it really depends and it's a case where a lot of engineering judgment and taste comes to bear. For example right now I'm maintaining a Jenkins system that has two large and complicated pipelines that are about 90% overlapping but for wretched historical reasons were implemented separately and the implementations have diverged over the years in subtle ways that now make it challenging to re-unify them. There is no question in my mind that this should always have been built as either a single pipeline with some parameters to cover the two use-cases, or perhaps as a toolbox of reusable components that are then used for the overlapping area. But I expect the mentality at the time the second one was being stood up was that it would be less disruptive to just build the new stuff as a parallel implementation and figure out later how to avoid the duplication.