> I've mentioned elsewhere in the thread that golden rice has yield problems, as do many GMO foods.

So do many non-GMO engineered crops where the trait that is the focus of optimization isn't yield, since a “yield problem” is deteemined by comparison to mainstream crops which are engineered for yield traits.

> That's partly because this kind of genetic modification isn't actually all that controlled

That's even more true of selective breeding, whether by traditional or more rapid modern means, so that's not a problem with GMOs compared to any real alternative, it's a problem of all existing methods compared to a hypothetical state where we have both perfect modelling to predict phenotypical effects of genetic changes and perfect ability to make arbitrary, error-free edits to genetic code. And even then we are stuck with the fact that some traits at inherently in conflict.