The rationale of the 83% rule (and the table in NEC2011 that preceded it) for dwelling service conductors is load diversity. I.e. the unlikelihood that a 200A service will operate all loads at once, or ever draw close to 200A. I feel the NEC's rules "double dip" the load diversity explanation, because load diversity is already considered in the algorithm that sizes the service in the first place, but that's a story for another time.
In any case, load diversity doesn't apply to a PV system, because it has the possibility of supplying the full nameplate amperes of the inverters. Even when you undersource the DC feeding an inverter, its output is still modeled as its full load amperes continuously. Therefore, it is best practice to limit the amount you can supply-side interconnect, to the ordinary ampacity of the service conductors. If building loads are temporarily zero, and the system produces full power, you will indeed put the full system current onto the service conductors. You'll ultimately transfer it onto the possibly-smaller utility conductors as well, but they aren't governed by the NEC.