First we have language in 690.14(C)(1) that says..."nearest the point of entry" and then we have language 690.31(E) that says..."first readily accessible location"
shortcircuit, with respect, you are carrying a simple misreading of the language to a totally unwarranted conclusion. The code simply does not say "first readily accessible
location", anywhere. The language in 690.31(G) is "first readily accessible disconnecting means."
In other words, you can take the 690.31(G) compliant method as far from the array as you like, as long as it ends in a disconnecting means in a readily accessible location. Whatever that location happens to be, it is 'the first readily accessible disconnecting means'. From there, if you like, you can continue it with a non-690.31(G) compliant method (not that I ever would, for reasons alluded to in this thread). Regardless of the safety theory behind this, this is what the code currently says.
As far as fire departments logic:
a) what Solar Pro said. How are firefighters going to find an inside disconnect on the upper floors of a building, especially a multi-story one? A
rooftop disconnect requirement makes sense. Any disconnect that is inside that is not at ground level makes little sense as a safety practice.
b) This argument is likely to be moot after the 2017 NEC comes out. Since rapid shutdown takes of it, the sections we are arguing about will be deleted from the 2017 NEC if the First Draft language sticks. An informational note will explain as much.
c) If firefighters are going to act that scared and ignorant of PV in a given case, then it makes little difference what disconnects exist, required or not.
...But I'm not sure that SunPower will let their integrator partners sell SunPower modules with a SolarEdge power conversion system.
Drifting totally off topic here, but ... I believe that the positive grounding required by Sunpower to prevent PID would make them incompatible with Solaredge and many other newer inverters. No?