The replacement of the axons by a shorter version that only represents the axon initial segment / hillock is a simplification for mostly computational reasons. Simulating the full axons would add a lot of compartments to simulation which increase the memory and compute requirements. Our current assumption is that not much computation is happening in the axon, and that the APs are submitted without much modification. Of course this is, as mentioned, a simplification and can be refined in the future. E.g. in the next release of the circuit we will include already some myelinated cable after the AIS to avoid boundary effects.
In practice, during the simulation when an AP is generated, this information is sent immediately to the relevant synapses of the postsynaptic cell with a delay that is calculated based on the path length along the full axon toward the synapse.
Werner Van Geit
Blue Brain Project