Correct. In a more narrow tube, where magnet fits better into the tube section, the effect is stronger. If the tube has a cut along its length, the effect becomes much weaker. The magnet must be moving for the effect to be there, so it never comes to a complete stop but may slow down a lot.

I recently completed building three dynamo hub wheels for bikes in my herd, a Swobo Dixon, a Big Dummy and a Steelwool Tweed I built up with a Shimano Alfine rear hub and drop bars. The Sanyo and a B&M Dymotec 6 will be going on lower use bikes.

Heck, it doesn't even need to be a tube.
I just reproduced this at home with a piece of aluminum angle stock and a neodymium magnet.

The amount of work that can be done is impressive. I'm pretty sure most of the free-fall style of amusement park rides use eddy currents as the primary braking force.

Eddy currents like this are used to separate aluminum from other recyclables. I'm sure if you look, you can find a youtube video. Basic process is spread recyclables in a thin layer, use conventional electromagnets to pull steel out. Then as the material goes over a drop, there's a rotating magnet. That repels the aluminum cans (and anything else conductive and non-magnetic) away from it. That's enough that the cans fall into a different place than the bottles, paper, and plastic. That increased level of mechanized sorting is one of the reasons that single stream recycling (where the end user can put everything in the same bin, and someone else sorts it out) has become practical.

That's what I thought too. I'm sure there's some tiny drag from the extra friction of extra moving parts, or something, but it's probably less than what you lose by not shaving that day.

Yes, but the question is what is the loss mechanism for the mechanical design and how does it compare? With a clutch there will be vibrations, slippage, etc which will lead to heat similar to the aforementioned parasitic eddy currents.

Keeping the magnets rotating serves the function of maintaining perfect contact and close tolerances. If you disconnect them then there either needs to be something mechanical to serve the same function such as a bearing OR you need to compromise the quality of the contact.

There's no free lunch. It won't be perfectly efficient no matter how you design it and, at a given price point, in all likelihood the more complex design will less perfect.

Obviously there are always exceptional examples. The point is that, given equal quality, a fixed design will be more efficient when on than a clutch which engages by friction. Just because the LightSpin had superior electronics (and was perhaps superior overall) it doesn't mean that the mechanical coupling employed wasn't inferior.