If people of some race are more likely to have a bad trait, but that increased likelihood can be completely explained by the fact that people of that race are more likely to have some other factor, then there is no correlation between race and that trait once you condition on the other factor. That’s what “completely explained” means.
If that increased likelihood cannot be completely explained by the fact that people of that race are more likely to have that other factor, then there remains some correlation between race and that trait even after conditioning on that other factor.
If people of some race are more likely to have a bad trait, but that increased likelihood can be completely explained by the fact that people of that race are more likely to have some other factor, then there is no correlation between race and that trait once you condition on the other factor. That’s what “completely explained” means.
If that increased likelihood cannot be completely explained by the fact that people of that race are more likely to have that other factor, then there remains some correlation between race and that trait even after conditioning on that other factor.
The problem is that in the grandparent the other factor is just a reformulation of “have the same race” in different words.