Once again, for AmigaOs, the major problem is really the amount of legacy cruft. As an open source developer, there are many constructions I would like to get rid off. However, as a project manager, I would certainly tell my developers not to touch them because they would break applications.
Forgive my ignorance, but to what actual closed-source development are you comparing open-source development of AROS?
Are you fantasising about legitimate, legal and commercial AOS 3.1 development as an alternative? If so, please explain why you think that would be an option. Without this option your pros and cons of open-source and closed-source development are totally irrelevant.
And, given the "competition" of 3.1 as we know it for more than 20 years, I have no doubts that any AROS68k developer interested in replacing 3.1 will be interested in maintaining not only source code compatibility but also binary code compatibility even though new developers will most certainly be discouraged from using some stoneage stuff. After all, without this binary compatibility AROS68k would be as useless for the vampire as some PPC-OSs derived from AOS principles.
