Obviously, this is purely hypothetical, but I don't see how imagining hypothetical solutions to hypothetical problems is beyond the pale. Every path begins not with the first step, but with someone imagining the journey ;)
Mimicking all this in silicon will be a formidable problem. You would have to copy all the turtles, all the way down.
Well, not exactly. I would suggest that you might be able to obviate all that by treating each neuron as a black box. You don't know what's in there, so there's no point in trying to perfectly replicate each and every aspect of a neuron. Rather, simply emulating how the neuron interacts with the outside world might be enough - I suspect that it's the way the aggregate collection of neurons interacts together that makes you "you", not something about the way each individual piece is made.
By way of an analogy, my quartz watch and my grandmother's hundred-year-old cuckoo clock are radically different from each other on the inside, and yet the core function of both - telling time - is essentially identical as far as everyone else is concerned. If I took out the gears and weights and springs from the inside of the cuckoo clock, and replaced it with silicon and a battery and an electric motor designed to act like a cuckoo clock, would anyone in your house even notice?
That could be done within limited parameters. But your emulation program might not respond to unforeseen circumstances the way neurons do, and probably neurons are as different from each other as snowflakes so you would need several zillion emulation routines to simulate a single CNS under controlled conditions. Possible.
They would notice unless you emulated the need to wind it, and emulated all the tempermental aspects of a weight driven clock.
More importantly, they would notice if those eccentric aspects of the original clock were somehow important.
Neurons are not like transistors that merely switch. They are electrochemical, and their switching behavior is profoundly dependent on a zillion chemical receptors that simultaneously modify their trigger level. This is all part of the "computation". You can assert that this can be emulated, but I remain unconvinced.
In addition to switching, neurons grow new connections with their neighbors and retreat from others. Connections are strengthened or weakened by whatever happens in learning. In short, we don't know much about what they do, much less how they do it.