I performed a test of the D800E and the Nikon 400mm f/2.8 lens with the oscillocope time base set to 1/2 second per division. Here is the result magnified by a factor of 6 in the vertical dimension:
Same second-order underdamped vibration. The time constant seems to be even longer than two seconds, more like three. The initial amplitude is about 18 pixels peak-to-peak (PP). To get it to one pixel PP, we need three time constants, or 9 seconds.
By the way, the head isn’t causing the vibration. I tried the Arca Cube and got the same results.