I think that is not a problem specifically of one field. DOIs, PURLs, and all sorts of indirection layers supposed to make references persistent are just technology solutions - when the real solution requires organizational commitment. It often turns out that at the point where DOIs need to be pointed somewhere else there’s no commitment left to do the redirections. Thus, we are likely to see DOI rot - hopefully of a somewhat smaller scale than the general link rot on the web, though.
I noticed that this week too. I it is the folks at linguist list who are responsible for that site. Contact them. I did find my copies of the GOLD ontology in the internet archive… The site doesn’t have any interactive features so if you download the latest edition with the comments included then you have everything there is to have. The dataset likely just needs a DOI via Zenodo. I could do that early next week.
As far as I know, LinguistList is in the process of figuring out how to deal with such legacy datasets. So minting a DOI now might create potential for confusion later. Also - as far as I can tell - GOLD was used in particular for Semantic Web stuff, so arguably resolvable PURL concept URLs for the GOLD stuff would be more important than just a copy of the schema file on Zenodo.