How would a device based on natural stochastic processes not solve boot entropy? I thought that was exactly how such devices were used now, except most users don't even have one.

I think the point is that cheaply manufactured devices based on natural stochastic processes are already a solved problem, and a new wacky process involving chemistry doesn't bring anything worthwhile to the table.

Here Hanno rejected all hardware devices:

> None of that is solved by any form of physics or chemistry based RNG.

Boot entropy? That is neatly solved by HWRNG.

Bugs/insecurity? These are indeed better with HWRNG. HWRNG are more likely to fail catastrophically (that is easy to detect if you don't use whitening) and easier to physically audit (there are much simpler options than this contrived crystallization device).

Algoritmic RNGs are more likely to fail undetectably and dangerously, and much harder to audit.