> E.g. how does this argument account for the data we have that someone with black sounding name will get less opportunity than someone with a white sounding name and an identical resume?

You're referring to a decades-old study that failed to replicate:

https://datacolada.org/51

(This is extremely common in social sciences.)