Hi Gary,
Thanks for the show and videos, that actually all looks to be working correctly, and I’ll attempt (!) to explain what you’re seeing here.
So you mentioned in the video that you expected the fader motors to be disconnected, and the behaviour is similar to this, but when you block anything using recall filters (Global/Scene) you’re blocking changes to the core rather than the faders. The physical controls are separate from the values in the core e.g. you can make changes to channel levels on a layer you’re not viewing with scene changes without any change to a physical fader.
The distinction is important because of the way the control (and feedback) of fader levels works.
So when you change layer, or mix, or change send levels for a channel you have on the current layer in any other way except by touching a fader - the level in the core changes, and also sends out a message to the fader to change position. The fader then moves to that position, and the reading from the fader is sent back to check it’s in the correct position.
If you then move the fader, it updates the value in the core, and again, sends out that message to everywhere else displaying that value (fader, screen, Soft Rotary etc…).
A good experiment to show how long this takes (not very long obviously) is to assign the same channel to two channel strips on the same layer and move just one of the faders. You’ll notice if you move very quickly, the fader that’s trying to follow will be just behind the one you’re moving.
The other part of this is that the SQ is constantly storing the current state of the desk. When you make a scene change with any filter in place, it must refer to this stored data to know the state it is currently in and then apply the scene change information (non-blocked) over the top.
So what you are doing in the video is moving the faders quickly and exactly at the point of changing scene. This means you’ve pulled the faders down part of the way when the core takes the reading of what the level is, and when it sends back out the message to tell the faders where to be, it’s sending out the level it stored from part way through your fade to -inf.
Best practice would be to either move the faders and then make a scene change, or make a scene change and move the faders, but not both at the same time.
You could also keep your fingers on the faders and they would take a split second to update to the new reading (-inf) but this would put strain on the motors as they would be trying to get back to that stored point for the split second they ‘think’ that’s the current state.
The small movement of the faders when at -inf is related to all this, as to increase the life of the faders, for tolerance and to avoid damage through overshooting (SQ faders are fast!) you’ll notice the -inf marking on the surface is actually a mm or so above the maximum length of travel. All the faders are doing is jumping to their minimum actual value (-inf).
Hope this explains it clearly enough, and enjoy the learning/playing with the SQ 
Cheers,
Keith.