We can learn a bit more about this dispute from how 23andMe has responded here.
Given that 23andMe is still willing to release their raw genotype results, it seems clear that the FDA isn't objecting to that level of their work. That's comforting: evidently at least that is considered more or less accurate.
Instead, they've only redacted "23andMe’s interpretation" of the data from their reports to new customers. My take on this is that the FDA wants some sort of rigorous demonstration that 23andMe isn't making errors in its interpretation of medical data. I'm thinking specifically of stories like this one: "My deadly disease was just a 23andme bug" https://news.ycombinator.com/item?id=6796219 It's hard to know (as a simple consumer) how common that sort of error might be: I rely on folks like the FDA to set and enforce quality standards on such things.
Similarly, I honestly don't have a great sense of how well 23andMe's interpretations reflected the full current state of known science. For instance, my memory is that they were only able to test for a few of the known breast cancer risk factor genes because others are still covered by patents. How clearly did their interpretation pages convey those gaps in what they were presenting?
I hope that 23andMe finds a way to get fully back online before long. But honestly, I'd be happiest if they manage to satisfy the regulators that they really are getting things right. Someone needs to verify claims like that for the rest of us!