My doctorate is in biophysics with deep experience in genomics. I've attended talks from the world's leading experts (as well as scientists/researchers/engineers) who work in this field, and read papers. From my inspection of the data, identity by descent does exactly what the researchers say, with a level of accuracy that is consistent with our current understanding of population structure and its changes in the past.

BTW, the main reason african genomic inference is so complicated is not the lack of genomic data, but rather that africa itself has far higher genetic diversity than any other continent, due to populations (whose genomes we can infer!) moving and interbreeding for a long time before humans left africa and spread out to the rest of the world.

(Article #1 is correct, there are definitely populations for which our current models are inaccurate due to bias and underlying complexity. I am not disagreeing with any of that).

(Article #2, where an individual compared the results, shows that the results are nearly identical, and well within the expected accuracy of the tests)

(Article #3 isn't interesting- yes, 23&Me and other companies have marketed truly terrible ideas. That doesn't mean IBD is scientifically invalid. The article also strawmans many of the realities underlying the nature of the tests and their relationship to disease prediction, or any other phenotype of interest).