I can't believe this is becoming another runaway thread, but here are a few points I hope will wind it down...
- Under the 2014 code, the 120% no longer applies to anything but panelboards. A feeder fed from opposite ends is not required to meet it, and a feeder tapped in any fashion in the middle is not allowed to take advantage of it. So this whole discussion of applying the 120% rule on top of a 'sum of breakers' for conductor sizing is simply not applicable under the 2014 code.
From this conversation, it sounds like the NEC seriously needs a definition and passage for dedicated PV system panelboards.
- Under the 2014 code, a PV combiner panelboard can take advantage of the rule that allows any combination of OCPDs for non-primary sources and loads as long as they don't exceed the rating of the busbar. While not explicitly defining and addressing 'AC combiner' panels, this rule effectively allows them to be sized at 125% of the inverter output (rounded up for breakers). Previous codes are admittedly much more ambiguous on this question. Unlike ggunn, I've heard of a couple AHJs requiring the 120% rule to both combiners and opposite-end-fed feeder conductors. Not for a while though.
- With regard to 3-phase combining, one could still read the 2014 rules as requiring that inverter currents or OCPDs (depending on the rule), be added simply rather than with vector math. For example, the AC combiner busbar in OPs example would be rated for 80A, instead of 63A. But this is a far cry from requiring the busbar or feeder to be rated for 119A. There's no situation under the 2014 code where the 'inappropriate' oversizing of any busbar or conductor would be by more than the ratio of 2/1.73. That's an oversizing of 15% instead of 91% (with a crazy strict AHJ). Again, previous codes are not so clear.
- Using three phase inverters avoids the problems we're talking about, and others. I hope to never have to combine single phase inverters.