In its last section, Example D3 presents three particularly interesting issues in regard to feeder neutral conductor sizing.
The first issue is load distribution. If you calculate the load based on square footage, you can assume it’s evenly distributed because even distribution is required [210.11(B)]. With an evenly distributed load, you are permitted reductions to the maximum unbalanced load [220.61]. Since you don’t have any assurance there’s an evenly distributed load, you must assume the maximum unbalance is equal to the full lighting load.
Example D3 then calculates that load, by dividing VA by V to get A. But why isn’t the ability to carry the return fault current a factor here? That brings us to our second issue: whether the neutrals are connected to the equipment grounding conductors (EGC).
The Example D3 feeder neutrals aren’t connected to the EGC, so they comply with 250.32(B). Example D3 says nothing about ensuring the raceways are continuous metal paths, so it appears that Exception 2 applies.
A third issue is whether the neutral is terminated on overcurrent devices. Given the particulars of this example, terminating on overcurrent devices is highly unlikely. Normally, you must size the conductor for enough ampacity to carry the noncontinuous load plus 125% of the continuous load (before any adjustment factors) [215.2(A)(1)]. But Exception 2 of that requirement allows you to size the grounded conductor as if the entire load is noncontinuous, if that conductor doesn’t connect to an overcurrent protection device.