The LED would also suddenly dim if the s-c was out of the circuit and the generator slowed suddenly. I'm not sure how the s-c voltage is managed, so I don't know if this happens. But as you say, if there isn't a decrease as speed decreases, perhaps the architecture you suggest is correct. I haven't noticed on my Luxos B.

I'm not saying dimming cannot be managed as you say or that it isn't managed as you say. I'm suggesting there'a another way of managing it than using software, which is to diode-OR the rectified and filtered generator output and the standlight energy store to a common point, or power bus, that has a higher voltage than the standlight. When the generator voltage sags, its switch (could be a MOSFET, a Schottky diode or a silicon diode) opens and the bus is fed from the standlight caps. Keep in mind I'm an old analog SMPS designer, so I still tend to think of analog solutions first.

But we are both imagining circuit designs. Neither of us will know how it is actually done without detailed investigation, and I'm not motivated to buy a family of B&M lights and dissect them, so this discussion should not go farther unless you want to collaborate on a design, perhaps an open-source LED headlamp controller design. Problem is, the real system benefits are in optics and light management rather than in power electronics and controls, so I'm not sure open source or DIY can improve significantly on what exists. I don't have any home optics software, at least, nor test equipment.

Thanks for the discussion.