You are perhaps confusing Cedar rather than Mesa which normally ran on the larger machines.
The other machines you mention were far too large and noisy to use in an office or normal work environment. NO ONE for example had a Dorado in their office, it was always in another room.
The Dorado came along later and played no part in Star, as knucklehead says. Dolphins were referred to as "dog-zeros" because they were so flaky.
Was he successful in getting it open sourced?
http://www.bitsavers.org/pdf/xerox/viewpoint/programming/610...
I worked on the Basic Workstation team; my part was the desktop and folders code. As I recall, the whole BWS rewrite was the result of Robert Ayers coming along and trying to rewrite the bottom layers of Star from scratch. The framework he came up with became the basis for BWS.
Traits was the object oriented multiple inheritance archetypes layered on top of Mesa that Star was based upon and required a static analysis step of all objects in the system to optimize object layouts—this is why it was a closed system. After the changeover to BWS, a few years later, only the document editor, and all object types represented in documents (graphics, tables, equations…) continued to use traits.
It would make the most sense to open source BWS (the basic workstation code used for the later version of Star) with all the tools. It had a ton of well written documentation that was all given to universities to use along with XDE.
A problem was that the InterLisp, Smalltalk, and Mesa development was done on the beefier Dolphin and Dorado (an ECL barn-burner) machines and as far as I can tell never tuned for the wimpy machines. I only used them when I tested some custom microcode on them; I did all my work on the faster machines too.