Has it been proven that people still manage to tell races of candidates apart after removing markers? Or is this all just conjecture?

Of course the "answer" is never that people were biased in favor of DEI groups in the first place, and removing said overt markers of their race just removed that bias because individuals could no longer discriminate. No, the answer was obviously then is that people found "secret" and "subtle" markers instead because they just have to discriminate and don't like DEI groups.

Occams razor comes to mind.

If "subtle markers" can still identify someone's race or gender, then remove those markers too. You can test of this works by giving employees bonuses to correctly guess the candidates' demographics and see if they can predict reliably.

If anonymization reduced the representation of certain demographics maybe it doesn't make discrimination worse, but rather you were wrong about which groups are discriminated against?

Do you have links to any studies that removing names and other obvious markers from resumes (college name, employment dates, etc) somehow increases discrimination in HR screening?

I honestly fail to how that could happen.

For example, if HR is throwing away all resumes that aren't from an Ivy League, then removing cities and schools from the resume can only help.