The potential difference between two points can be calculated along any path between them. I take a parallel plate capacitor and consider a small positive charge on the surface of the negatively charged sheet. I move it outside the sheet without doing any work as the net field inside a conductor is zero. Now, behind the sheet, the fields due to the two plates cancel each other perfectly. I move the charge away parallel to the plate, infinitely far away, when any fringes would've decayed away to nothing and move it to the other side. I repeat, bringing back the charge to the other plate. The two plates are seemingly at the same potential when they clearly are not. At what point of my process did I do work? What exactly is the potential at all points in space due to this system?