Yeah I get your point. But I guess for this model you can kinda have a concept of the "ideal" training set, where all high frequency features appear at the same rate as in real world.
I don't understand how people can defend his detractors in this particular case. Are you telling me that an image upsampling model that does not contain hard coded bias, and trained on unbiased data will produced biased result? Especially the kind of biased result represented by the error made by the original tweeter who fucked up?
Large HIV database inflating matches is indeed a big concern. But dismissing one miss matches sounds arbitrary: these segments were not arbitrarily selected, but real insertions on tops of sars.
Could easily be "mixed samples" on the other end instead of intentional leaks. You know how old lab equipment could be really useful in a live seafood market.