A potential answer to this occurred to me while reading your question, specifically your section about when demographics should be used:

In scientific research, demographics are only requested if there is evidence that a demographic attribute will lead to a significant difference in responses

Perhaps this is what SO/SE is trying to reveal, is there a difference in user satisfaction based on race/gender/age?

Perhaps 75% of white men will report a positive experience on the site, while non-binary individuals report only 40% positive experiences.

I am in no way defending the way the survey was done or the lack of appropriate diversity in options, but I do see the above scenario as a possible reason for demographics to be present in the survey.

A potential answer to this occurred to me while reading your question, specifically your section about when demographics should be used:

In scientific research, demographics are only requested if there is evidence that a demographic attribute will lead to a significant difference in responses

Perhaps this is what SO/SE is trying to reveal, is there a difference in user satisfaction based on race/gender/age?

Perhaps 75% of white men will report a positive experience on the site, while non-binary individuals report only 40% positive experiences. (Please note these two groups are not mutually exclusive.)

I am in no way defending the way the survey was done or the lack of appropriate diversity in options, but I do see the above scenario as a possible reason for demographics to be present in the survey.