It would seem to me that the only way operation of a heated AI surface can lead to the formation of ice aft of the protected area is if the protected area is "running wet" as opposed to "fully evaporative" i.e. there is (only) sufficient energy to raise the temperature of the supercooled droplets such that they are simply droplets, and run back along the protected surface, only to freeze in contact with the (colder) unprotected surface further aft.
In such a scenario, turning the system on and off would seem only to further guarantee that in some circumstances the running wet condition occurs - as the protected surface heats up or cools down, its almost certain to pass through the running wet case, especially if the full power case is also running wet. So I fail to see any practical gain to be had - the same risk is still present.
If the protection actually is fully evaporative then cycling is worse, since it exposes you to the heatup-coooldown cycle and a possible running wet phase that is not present in full power operations.
Additionally, Airbus will (I am 100% sure) have demonstrated the performance of their system under some quite extreme conditions and developed procedures accordingly. If they did get runback ice, it will have been accounted for. Compromising the protected zone (as any cycling must do) to protect against a threat that's either theoretical or, if real, already addressed, seems like a poor option.
Disclaimer - "our" systems are fully evaporative and and anyone suggesting such a procedure on our types is being, at best, a bloody idiot.