I agree with you. Of course, sorry.
Constant 0 V is expected in V with a short-circuited input. Regardless of the vert. shift. However, V is not constant and varies with Y displacement. With this behavior, you don't have to worry about temperature-dependent offsets.
I thought a time marker was unnecessary because it affects the whole video.
Attempting an explanation:
If the acuracy related to the y-axis in the display area is 8 bits, then we have 10V/div x 8 div = 80V. And 80V / 256 = 0.3125V as quatization. With +1Lsb Toggle (due to bad calibration), Vmean can of course now wobble between 0...0.3125V on average (calculated as a float). As observed.... But then Vmin = 0 and Vmax = 0.3125V would need to be. But it is not.
Maybe someone has an explanation and it's not a bug but an unqualified expectation on my part...