Take a look at this other thread: https://www.electrical-contractor.net/ubb/Forum1/HTML/006324.html
Kinda of funny how the same topics can come up multiple times in different locations.
One of the points that I was trying to make in the other thread is that you don't need _exact_ balance in order for the neutral to not _count_ as a current carrying conductor for 310.15(B)(4).
Perhaps a better way of looking it is not to ask 'does the neutral carry current?', but instead to instead examine the sets of 2, 3, or 4 wires comprising a circuit and ask the following two questions: 1) Do any of the individual conductors carry excessive current? 2) How much total heat is generated by this _set_ of wires when loaded in the worst possible heating condition?
In those situations where the code permits the neutral to not _count_ as a current carrying conductor, you will find that the worst case heating for a given set of N conductors happens to be equivalent to N-1 conductors fully loaded.
In a 3 wire _single phase_ multiwire branch circuit, the worst case heating happens when 2 of the 3 wires carries full current, and the other carries no current. You can get this situation with one branch of the circuit fully loaded, and the other branch unloaded, or with both branches fully loaded. You might have one hot and one neutral fully loaded, or both hots loaded with the neutral unloaded. But in either case the total heat production is maximized. In the case where the current is _not_ fully concentrated in 2 of the conductors, the _total_ heat production is reduced, even though all three conductors are actually carrying some current.
Consider the above case where the resistance of each conductor is 0.1 ohm (roughly 50 feet of #12 wire), on 20A circuits run on 12ga wire. In the case of both circuits fully loaded, you have 20^2 * 0.1 = 40W being dissipated in hot A, and 40W in hot B. In the case of one circuit fully loaded you have 40W in hot A and 40W in the neutral. But in the case of one circuit loaded at 20A and the other circuit loaded at 10A you get 40W in hot A, but only 10^2 *0.1 = 10W each in both the neutral and hot B. So even though all three conductors are actually carrying current, the heat produced is less than if only two conductors were carrying the current.
The same analysis will hold for a 4 wire three phase circuit with nice resistive loads. The worst case heating for these for conductors is the same as for 3 conductors fully loaded.
In the case of a three wire three phase circuit, the neutral ends up carrying almost the same current as the phase conductors, so the worst case heating for these three conductors is the same as three conductors fully loaded.
For a 4 wire circuit with non-linear loads, the worst case heating is harder to figure out, and in some case is _worse_ than 4 conductors fully loaded.
For a 3 wire single phase circuit with non-linear loads, things are a bit different. Most of the non-linear loads used today produce _odd_ order harmonics....which happen to balance on a single phase circuit, and don't load the neutral. So loads which would heat the shared neutral in a three phase feed will not be a significant problem for the shared neutral in a single phase feed.