Press pause on a VHS deck and the tape stops, but the head drum does not. It keeps spinning at 1800 revolutions a minute on an NTSC machine, and the heads keep sweeping across whatever piece of tape is now sitting still in front of them. That is the whole problem: the tracks are recorded as long diagonal stripes, and a head that is not moving along the tape no longer follows one stripe. It crosses them. Crossing them is an issue because of a trick VHS uses to pack tracks tightly. The two heads are tilted, one about six degrees one way and one six degrees the other, so that each head is nearly blind to the track its neighbour wrote. The short wavelengths, which is where all the detail lives, cancel out across the width of a track read at the wrong tilt. That is what lets adjacent tracks touch with no guard band between them and no crosstalk to speak of. When the head crosses onto a wrong-tilt track, the same trick works against you. The signal coming off the tape collapses, and the luma on a VHS tape is frequency modulated, so what the demodulator has left is no carrier at all. A frequency demodulator with nothing to follow does not output black. It outputs whatever it happens to lock onto, sweeping the full range from below black to above white, which is the same thing that happens over a dropout: random black and white streaks. So the band is not a shadow or a smear. It is a strip of live static, and in the clip you can see the difference between the two parts of the picture: the picture above and below it is frozen, while inside the band the noise is different in every frame. Measured on this render, the band changes by about 26 levels of brightness from one frame to the next while the held picture changes by 3. Where the band sits is not fixed either. It depends on where the drum servo happened to be when you hit pause, which is why pausing twice in the same scene can put it across someone's knees one time and across their face the next. Our model picks its position the same way, once per pause.