Agumander

With their massive inertia, elephants would probably change station attitude every time they pushed against a surface.

Perhaps a mass-driver station could have big engines on it for station-keeping? It would then require periodic refueling missions, but then that fuel wouldn't need to be carried by every craft that wants to make the voyage.

Doing jumping jacks as the rocket accelerates is a bad idea, for the same reason you don't want to be standing on a skateboard on a subway car as it changes speed. If your weight isn't part of the load, then you aren't being accelerated. You'll find yourself quickly hitting the back wall as the vehicle accelerates around you.

Normally mouse and keyboard, but I also have a Steel Battalion controller (and pedals) that I've been trying to get working. It already works well for some other games, but not with KSP or other Unity games since the engine ignores system calibration settings. I might try writing a plugin to get it working, which will let me access all the buttons instead of just a few. I should also be able to make them blink in response to in-game events.

Is there anything special about the manufacture of the test device such that a mechanically-inclined person couldn't build it in their garage? It seems like a resonant cavity shouldn't be that hard to build... but I guess there's a reason the magnetron from a microwave oven won't work as an emitter for an EM drive?

It could be done with just one camera. The "pointiness" could be determined from the surface normal, which is already given in Unity's rendering pipeline. Side cameras would miss any indentation in the front surface, but these would be visible to a front camera. The drag could then be scaled by the dot product of the surface normal with the drag camera's forward vector. I'm not sure offhand whether an indentation on a front surface would produce more drag than a totally flat surface. Either way, to accurately model drag this way they'd need to be detected. Let's say the front camera uses a shader that includes calculation and transformation of the aforementioned dot product. Indentations would appear to the camera as regions outlined in high drag containing regions of lower drag. A modified flood fill algorithm could be used on these areas to represent the drag of the indentation, as long as the method for differentiating between the inside and outside of these regions is reliable.