I have to wonder if there is a bad connection and there may be some arcing going on. If you have an infrared camera, might take a look at the connections and components in that leg. Seems to me that a power engineer might have some insight. You describe all of the tests that you did, but not the results. I’m assuming that the problem stayed with the cable.
You may have something there. There’s a graphic description in the OEM Learn section on AC Power Theory that illustrates how load imbalance between the phases can shift the angle between the phases:
https://learn.openenergymonitor.org/electricity-monitoring/ac-power-theory/3-phase-power
To the extent that voltage is equal on each of the phases, apparent power (VA) should be the same, but you’re right, the real power calculation is affected by phase shift, as is the PF, and that could be happening, but I can’t get my head around how it could affect just one of the three parallel cables and not the others.
This two cable setup is looks pretty consistent. Would be nice to have at least one big CT to validate each leg in turn. There are some pretty good deals on that kind of old industrial stuff on Ebay. You would need up to 50ma output at full load.