-snip-
I've read through the links you provided, and while there are some good things about their system, there are some bad things as well.
I think there is no ideal approach to achieve 100% randomness using at least one computational device (in their case TRNG) no matter how to mix entropy coming from various sources.
But,in my view, the scheme used by NGRAVE is worth sharing.
No doubts, your analysis of their techniques is vary valuable and should be taken into account.