It's possible that the neural scanning process will need to be destructive in order to work properly. Mapping the neural connections isn't enough; you must also capture enough state that each copied neuron responds as the original neuron does. The task is to get a high resolution, three dimensional snapshot of a highly complex organ that's changing even as you're trying to make the recording. If you take too long to make the recording, brain states at the end of the session will be wildly divergent from the ones recorded at the beginning and the copy will wake up with a scrambled consciousness.
A blackbox approach to copying such as the Moravec Process embraces the inevitable changes in brain state during recording by probing the brain from the outside in, modelling neuronal behavior statistically, learning how each neuron responses to stimuli. When a neuron's behavior has been learned its behavior is imitated by the probe and the neuron is cut away. The probe reaches deeper into the brain and repeats the process. At the end a working copy of the brain exists, being fully imitated by the probe, but the original is utterly destroyed.
If you decide to assay internal neurochemistry without the fancy interactive nanoprobes, then you have to stop the brain from changing while you're recording it, else you're back to having the copy wake up with scrambled brain states. So you stop the brain from changing by killing it quickly, for example by freezing it, then you get out your electron microscopes and microtomes and go to work. You'll need software to correct for damage from swollen and ruptured cells, inevitable stress cracks due to parts freezing at different rates, chemical changes due to freezing, and so on. But you've already perfected the procedure on dogs and great apes beforehand, so the software knows what to do. In the end, you have a working model of the brain inside a computer, but the original brain has been chopped to bits and has no hope of reassembly.
