Not just mutations. The MUCH bigger factor would be a child whose phenotype is "desired" but whose genotype is "less than ideal," or simply a kid who was the best they could get with the limitation of the parent's genetic material. Like, if AA = bad, AG = good, GG = better and the parents are AA and AG, oh, well, let's just make an AG kid.

You can't just make one generation of genetically filtered children and figure everything's going to be good now, we have all the "bad parts" eradicated, because that's not how genes work. In addition, if you read the paper, it mentioned something on the order of a dozen embryos, and doing a genetic analysis of each one. Using this technique, you'd be able to filter out, perhaps a handful of disease, maybe pick out one with green eyes if that's what you wanted (and your genes allowed), but you're not going to gene-by-gene eliminate every chance of bad combinations with every other combination of genes on the planet, barring mutations. That's ridiculous.

And how would we know which mutations are broadly beneficial or not? What if humans of a future earth or larger society have needs we couldn't imagine, like radiation resistance or anti-autoimmune protections?

Besides, sometimes there are trade offs with no clear winning picture. Sub-Saharan Africans with sickle cell syndrome are also more resistant to malaria. What's worse? What's more necessary? How do we know malaria won't be a bigger problem tomorrow?