To provide a counter-example, hydralazine and isosorbide dinitrate are combined into a single dose as a treatment specifically for black folks with heart failure (initially sold under the brand name BiDil), because the combined drug treatment in general works measurably better on black folks than white folks, to the point that the combo drug was rejected by the FDA based on initial trials (that had a majority white patient base), but was later approved specifically for for black patients because that specific pair of drugs worked enough better in that population to be approved after further trials. It's fallen somewhat out of use as a treatment, not because it was ineffective or "racist" to approve a race-specific treatment, but because better options have been developed in the last 20 years - the drug combo remains approved specifically for heart failure in black folks, however. It's just no longer the first choice.
common social categories of race (scientists use ethnicity, because of eugenics),
Literally, they use ethnicity because of negative political associations with race as a term, and also because from a practical standpoint ethnicity is like race, but with more narrow groupings in modern parlance (as noted in the past "race" referred to much narrower groupings, closer to how ethnicity is used now).
Also, eugenics would totally work if we weren't terrible at deciding what "good genes" are and instead inevitably make it about something dumb like skin color and there weren't the massive ethical issues in actually doing it.
Here's a fun question: If you had to choose a hypothesis that would be functionally impossible to properly test because of ethical or political issues but that you strongly suspect is true, what would it be?