The maneuvers are so extreme and come so fast that I would not have been able to say for certain that this wasn't just a very nasty crash in progress. But they were, in fact, completely controlled and intentional.
Incredible.
See https://en.wikipedia.org/wiki/On_Being_the_Right_Size (PDF at https://www.phys.ufl.edu/courses/phy3221/spring10/HaldaneRig...)
The tests that have shown "significant" improvements have frequently compared the Sharrow to a sub-optimal prop. Feedback from many actual users is that the gains are moderate over a narrow RPM range.
The other thing I was thinking of trying is swapping in a different "high torque" lower unit with a lower gear ratio and running a significantly larger prop.
A real 3D aircraft, however, has a fuselage. Similarly, a prop has a hub and the tips of each blade are spinning faster than the roots. The tl;dr of this is that real 3D lifting surfaces typically exhibit a mixture of chordwise and spanwise flow, which causes wingtip vortices to form[0], resulting in induced drag/induced power loss.
For a given amount of thrust the total amount of momentum that the prop transfers to the fluid is fixed. The tip of a conventional prop ends abruptly which causes a large pressure gradient and a strong vortex. A toroidal prop's shape causes the pressure gradient to be broader and less concentrated, therefore the wake vorticity is distributed over a larger region, reducing peak swirl velocities and lowering the kinetic energy lost to vortex formation (and to cavitation).
If I remember right what they didn't do was go exactly straight. You could see a (very modest) s-shape in the wake over distance.
ref: https://www.shipsofcalmac.co.uk/fleet-features/the-streakers
Most traditional tugs have a pair of screws for just this reason. Not so much to turn but by applying differential thrust they can pull sideways. A vector drive like this will vastly increase the envelope of possible pull conditions.
