Tightening the nut partially corrected it.
I think the clutch nut and circlip both contribute to eliminating endfloat. Let's look at the drawing and imagine the circlip totally fatigued, not just partially. It is so fatigued that there is a gap, between the circlip and bearing, even with the engine stopped. That would be worse again, possibly with a knock at all rev ranges. In that case you would still have a knock even with the clutch nut at 150 ft/lb.
On the occasion that the noise remained after pulling the clutch in, until it got to 20km/h, that again points to the circlip, I think. Pulling the clutch in compresses an already weakened circlip. In the video I think I hear the noise momentarily between gears when the clutch is pulled in. That was why I asked earlier about pulling the clutch without disengaging it, since it only happened under load.
There's a service bulletin relating to the earlier version about output shaft noise, due to, amongst other things, a loss of pressure applied by an idler gear diaphragm spring. The minimum action recommended, was that it be replaced. In the newer version, it's replaced with shims, pressure is applied on the bearing by the circlip. That also makes me think that replacing the circlip could be a fix, and can be replaced without splitting the engine, so that's one good thing.