I understand that and I really like this UX.
For now I'm using the published state to get the latest result available (artifact). I'm using it wrong I guess and should use tags or dates instead. But it helps me to have only one experiment tagged as published which would be the only ground truth (and if a new one comes it takes that role)