@fair ,
I see one major flaw in your logic. CD and two channel DSD is just that, two channels. When I am in a room, out in the wilds, or anywhere, there could be an infinite number of sound sources, that all contribute to that data you mention. When I am at home, there is only 2 sound sources. They may bounce off the walls, the floor, the windows, but there is only 2 sources. In another thread we are talking about ATMOS with 9, 11 or more speaker which still only simulates all that we can hear.
Use that 11 speaker example at CD data rates. The rate is 7.8 mbits / second. 11 speakers is not enough. 24? Now 16.8 mbits/second. Well beyond your 3.5 - 4 mbits/second.
I don't think you can correlate the data rate for the cochlea with the brain, which I suspect is a WAG, from sound information that comes from all directions, with what comes out of 2 speakers.
Auditory processing circuits in the brain drastically compress this flow of information: this explains why lossy encoding works so well. Still, if some part of the original 3.5. to 4.0 MegaBits per second flow is arbitrarily removed, artifacts may occur.
As we can see, CD falls about 5.7x short of the target of complete digital transparency. DSD64 falls about 1.4x short.
DSD128 encodes more information, by factor of 1.4x, than the nerves running between cochlea and brain can transport. DSD256 exceeds the sufficiency threshold by factor of 2.8x, and thus shall be considered far more than enough.