I absolutely agree that there are 4 current carrying conductors, and that for purposes of derating they should be treated as 4 current carrying conductors. They physics is pretty clear.
I brought this up because I was pondering the _actual wording_ of 310.15(B)(4), and trying imagine situations where the wording 'falls down'. After pondering 'idealized extreme' cases where the wording fails to match the intent, then I can ask the question: are there real world situations where this would matter?
I have three ungrounded conductors each carrying 24A. I have a grounded conductor acting as a return for the _net_ current being supplied by the ungrounded conductors. I believe that there is a strong argument that under the _wording_ of 310.1(B)(4) this conductor is not counted, even though I totally agree that it should be counted.
While my example above is not a true MWBC, since the single phase load is entirely separate from the line to line load, it is the extreme case of a MWBC that happens to serve line to line loads as well as line to neutral loads. MWBCs are permitted to service line-line loads. A circuit that has 24A from phase A to neutral, 1A from each of B and C to neutral, and 23A from B to C _would_ be a MWBC, but would have very similar physics as the extreme example. All 4 conductors would be loaded in excess of 20A.
What about the following changed example: Given: 120/208V three phase wye service. A 30A 'full boat' (three hots and a neutral) in a conduit. All loads are 120V between hot and neutral. The loads are phase A: 120V 24A resistive, phase B: 120V 24A 30 degree leading power factor, phase C: 120V 24A 30 degree lagging power factor. No harmonics at all, just power factors caused by inductive and capacitive components. (To be clear, I am trying to set up a MWBC with return on the neutral where the actual current flow is the same as my example above.)
The real world situation that might apply is the neutral loading on a feeder to a panel that serves both line to line and line to neutral loads.