The delayed choice quantum eraser experiment works too, but I was thinking of a two-channel (CHSH) Bell test where the observations are made separated by a few

The delayed choice quantum eraser experiment works too, but I was thinking of a two-channel (CHSH) Bell test where the observations are made separated by a few light-minutes, then later compared with each other (and somehow you eliminate enough noise that you don't need a coincidence monitor).