I know how ancestor disagreement works. The observation was previously at class for this reason.
However, a recent change was made to the way ancestor disagreements are calculated, so that it is not treated as a disagreement with other IDs in the same family once the specific ID that was being disagreed with has been withdrawn.
That is why the ID is now being displayed at species level once I clicked and unclicked some DQA to get iNat to reindex it.
But if the disagreement is no longer a disagreement, it should only take 2 IDs to get to research grade.