Excellent classical reference!
I think getting standard, clean, accurate metadata given the huge number of titles and the ever increasing number of releases of each title is a nearly impossible task. I don’t see any software startup having even close to the resources to get it done, and there will always, always, always be another title that hasn’t been cleaned and filled out.
Crowdsourcing might help. It would also create an entirely new dynamic within the Roon software ecosystem that may not be part of their business plan.
My expectation is they will try to do a sort of Big Data approach and triangulate on the best metadata using multiple large sources. It will never be perfect but it will improve over time, assuming Roon is able to cut affordable deals for the data.