HMMM ok. I agree it is a garbled mess. So the question is, I am just trying to follow the code path for each of the following situations: tramsformerless inverter, transformer inverter, and micro inverter (which I know doesnt have to be its own thing, just bear with me).
First, the technical distinction is between grounded and ungrounded DC systems. Enphase's latest line of inverters, for example, are ungrounded but also have a transformer in them.
My understanding is that the CMP accepted the argument that because UL standards don't require a GEC for an ungrounded inverter, the code needn't require a GEC from the PV system. The entire PV system still needs have equipment grounding which is ultimately connected to earth, but there doesn't need to be a conductor(s) between the inverter and electrode that meets all the requirements of 205.166.
If you ever get the chance, put your clamp meter around the GEC for a larger grounded system: you'll see a bit of current flowing. Not so much for ungrounded systems. Has to do with module frame capacitance, IIRC. So, yes, there is some real reasoning behind a slightly larger, more bonded GEC on grounded systems.
As far as how garbled it is, check out this interesting tidbit from the 2014 ROP:
The change to 690.47(C)(3) is in response to the TCC Directed Task Group
consisting of CMP-4 and CMP-5 members to discuss the conflict between
250.121 and 690.47(C)(3). The change in 690.47(C)(3) from the term
“Grounding Electrode Conductor” to “PV Bonding Conductor” resolves the
conflict with 250.121. The other changes reflect the change from a grounding
electrode conductor to a bonding jumper.
250.121 says that an EGC can't be a GEC, but 690.47(C) says they can be combined.
1. Ok so the CMP obviously stayed an extra day just to make it as unclear as possible, but I sort of follow how to arrive at the GEC requirements for transformer inverters, but where does the code split and tell me/not tell me that I need a GEC for a transformerless inverter
I'd say it doesn't split that way, which is indeed confusing.
The last sentence of 690.47(C)(3) says you can size your GEC for your ungrounded inverter to 250.122. One notable AHJ I work in takes this to mean the whole PV system just needs to be bonded according to 250.122, which I don't have a problem with. OTOH, one interpretation I've seen (in a Solar Pro article), is that the GEC can be smaller but still has to be installed as in 250.64, i.e. unspliced or irreversibly spliced, bonding bushings, etc.. The best argument against this is that the bits in 690.47(B) say that ordinary EGC bonding is permitted.
2. Why do microinverters (I know at least some are isolating) not have to have an GEC? Maybe they are using the Combined EGC/GEC method, but then how do they get around the "irreversible splice" clause?
There's no particular rule for micro-inverters, which aren't defined in the NEC. Enphase's older models had a grounded DC conductor, and thus required a GEC, which their manuals made clear. However their newer models are ungrounded, and the manuals no longer mention a GEC. Plus, they published a whitepaper pointing out that UL said they no longer needed a GEC, which swayed a lot of AHJ's for me.
What is your explanation of 690.47(D)anger?
That's really an unrelated subject. But since you ask, I put it down to three factors: a) confusion over bonding vs. grounding, b) people forgetting that the scope of the NEC does not include preventing/redirecting lightning strikes, which is covered by a different code, and c) a North American fetish with grounding.