For the variability point, they do tests in as a controlled environment as they could, and do the tests until they get consistent data. But what do you mean by significance test and how can they do it?

I agree they're not the gold standard but they're the best we got in terms of independent third party testers, and I would assume they're more than good enough for tech stuff.