Since it is a transformerless inverter, the array conductors which are not explicitly grounded will end up referenced during operation to one or both of the AC output terminals of the inverter. That means that the voltage to ground of one of the wires of that subarray at any given moment could be the DC output of the array plus or minus the instantaneous voltage at one AC output terminal.
Same applies to each array.
Now if the design is such that one DC lead will be near the voltage of on AC lead and the other will be offset in the direction of the other AC lead by the Vmp voltage, the result should be that the voltage on the other DC lead will still be less than Vmp from ground.
But I am not convinced that you will not potentially have voltages between two conductors of different arrays that are more than Vmp if the two inverter's AC outputs are on different AC phases.
As long as each conductor is insulated for 2 x Vmp, that should be OK. I just feel bad about it.
I am not at all convinced yet that the voltage between any two leads will never be more than Vmp
With a grounded non-bipolar array you can guarantee that relationship.