Using some actual hypothetical numbers:
Assume a trunk cable rated for 20A and a 20A breaker. This can support 16A of total module output (125% for continuous current).
Now assume that temperature correction at rooftop temperatures reduce the ampacity of the trunk cable to only 12A.
If I place 8A with of modules on each of two runs of trunk cable combined (without additional OCPD) into one 20A branch circuit there is no way that the inverter outputs can overload either trunk cable segment.
But a resistive short circuit at one of the modules (or a skinned cable in a water filled conduit) could theoretically draw 20A of fault current through the breaker indefinitely, overloading one 12A segment.
However if you are willing to accept this argument against this configuration you would also have to disallow 16A of modules on a 20A circuit with no derating because a hypothetical resistive fault could pull 20A from the breaker and 16A from the modules, causing a current of 36A near the fault.