GFP revisited for 2017 NEC

[ICPete]

Happy New Year to all you PV and electrical wizards!

I'm in the final stages of designing my PV system and will be getting the schematics reviewed by a state-licensed PE, which is required in my county because I'm self-installing (long story why no luck hiring solar contractors).
I bought a copy of the 2017 NEC as well as Bill Brooks and Sean White, "PV and the NEC". Many late nights reading these tomes. Brooks and White is extremely helpful.
I believe Bill Brooks may be the "BB" who has posted here and on the AZ Wind & Sun forum regarding the hazards of DC GFP (at least 8 years ago!). I've read most of that discussion as well.

My design logic is as follows; I'm looking for feedback as to whether I'm looking at this correctly, whether the code should actually be followed or ignored as far as GFP for the Classic PV strings, and whether I'm missing anything:

1. Since 2008 the NEC has required GFP on building-mounted PV strings. It needs to detect ground faults in the PV array circuits (including, for 2017, any functionally-grounded conductors) and either disconnect the conductors or stop the operation of the charge controller (or inverter as the case may be). If the latter, it must also isolate the PV system DC circuits from the ground reference in a functionally-grounded system (this clause is new for 2017).

2. The Classic's GFP implementation will only detect faults that short the + PV conductor to safety ground; obviously it can't detect shorts between the - PV conductor and safety ground, as that would simply be a short in parallel with the PTC. I believe it's possible to design a circuit that WOULD detect both polarities of ground fault, but AFAIK, the Classic is still using the PTC approach.

3. As well, the Classic's GFP implementation shuts down operation of the charge controller (which was sufficient in 2014 NEC), but does NOT isolate the PV conductors at all. So it doesn't do either of the two allowed methods listed in 2017 NEC 690.41(B)(2). I'm not trying to criticize the design of the Classic; it's a fantastic charge controller from everything I've learned about it so far. But from my reading its GFP implementation doesn't appear to meet the requirements of the 2017 NEC in 690.41(B).

4. Since the battery - conductor is solidly connected to the Classic's - conductor, it follows that the negative battery lead must not be solidly grounded either, if GFP is enabled in the Classic.

5. 2017 NEC 706.7(A) states that a disconnecting means shall be provided for all ungrounded conductors derived from an ESS (energy storage system, e.g., battery). Although Article 706 only applies to ESS operating at over 60V DC, my 48V nominal battery could conceivably go slightly above that level during end-of-charge conditions. So this would imply the need for a 2-pole breaker, or separate 1-pole disconnects functionally tied together. Article 712 has a bunch of very specific requirements for DC microgrids, the definition of which sure seems to include our battery bank, PV arrays, and charge controllers (DC-DC converters). This article clearly requires disconnecting means for both conductors if the negative is what they call "reference-grounded".

6. The MidNite E-Panel architecture doesn't appear to be designed to disconnect both conductors from the battery at either the battery inverter or the Classic.

7. If we ignore for the moment that the Classic's GFP only detects ground faults in the + PV conductors, it seems we could make use of the programmable auxiliary relay to trigger additional disconnecting devices. I'm thinking of multiple inexpensive vacuum contactors, triggered by a latching circuit so that a +12V pulse from the Classic's auxiliary contacts opens a normally-closed relay, shutting off the multiple vacuum contactors, with latching logic to keep them off until the system is manually reset. I could place a pair of contactors in the + and - conductors running to each battery inverter, as well as a pair to disconnect the + and - conductors running to the Classic(s). I'm planning to use these vacuum contactors for remote disconnect of the battery circuits anyway (for safety shutoff); this would just double the number I need to get.

So I think this scheme addresses some of BB's concerns regarding DC GFP using a fuse or PTC to break the bond between the negative conductor and the safety ground. When the GF is detected, the battery bank gets completely disconnected from everything else. Heck, I could add several more vacuum contactors (these could be small ones) to disconnect the arrays from the Classics, and could even install more contactors to short the PV arrays at locations up on the roof (probably would need Normally Closed contactors in this role). But actually, since I plan to install IMO FireRaptor RSD units at each PV module (or one per pair of modules), I could simply use the GFP auxiliary signal from the Classic to shut down all the PV modules supplying the master and slave Classics.

The vacuum contactors I'm looking at are Gigavac units that are UL508 Recognized and load-break rated.

The alternative, of course, is to NOT enable GFP in the Classics and instead to solidly connect the safety ground to the battery negative conductor (and therefore to the Classics' negatives and to the PV array negatives). This sounds like a safer approach, according to what I read from BB's several postings on the subject. I will have to ask my local AHJ whether they will insist on GFP for the PV circuits, or perhaps they agree that it is less safe.

So what am I missing in all this?
What GFP approach is being taken by others installing Classic charge controllers in jurisdictions that follow the 2017 NEC?
Are people biting the bullet and installing two-pole disconnects (breakers) all over the DC bus?