I disagree with this one. If the load side circuit is ultimately protected against the current adding up as you described, it should only need to be sized per the maximum of the two sources. Not the sum total of both sources.
Here's a sketch of an example:
View attachment 2554385
Worst case scenario, the loads draw 200A, regardless of what source feeds them. Any greater than that, and the main breaker of the main panel would trip. Any breaker in this diagram, could alternatively be a set of fuses, and the same reasoning will apply.
At points A and B, the current Ia and Ib will be no greater than 200A, otherwise, the main panel's breaker would trip.
At points C and D, the current Ic and Id will be no greater than 200A, otherwise the service disconnect will trip
At point E, the feeder tap of interconnection, the PV OCPD limits current Ie to 80A. This really corresponds to the inverter limiting it at 64A or less, but that's a detail for another time.
If Ie = 80A and Ia = 200A, then Ic = 120A. The PV subtracts current from the main supply, so that only the remaining current comes from the grid.
If Ie = 80A, and Ia = 0A, then Ic would equal -80A, indicating that we are exporting 80A to the utility.
If Ie = 80A, and Ic = 200A, you could not have this happen without a fault or an additional load.
In order for loads in the downstream panel to draw 280A, you'd need this to be an MLO panel without a 200A OCPD between it and the point of interconnection. Then in that case, you'd need to upgrade the main panel amps to meet or exceed 280A (which would mean a 400A busbar). It is statistically unlikely that this would happen (given how conservative load calculations can be), but still a possibility that requires protecting against. Alternatively: you'd need to connect on the supply side of the service disconnect, so nothing downstream of the point of interconnection draws more than 200A.