As I understand it then, the issue at the RCD is that even though the mains supply is AC, I could apply a pulsed load. If (for instance) this load is always near the + peak, the effect is a "nearly DC" current. This DC bias current can saturate the RCD coil.
To measure this effect in my own device then, I'd have to provide equivalent response to a type 'B' RCD. The type B devices work by measuring the effect of DC bias on the inductance of a supplementary sense coil. It might be the only sensible option is to copy this technique (?).
Given the existence of type B RCDs, why do some end-user devices on the market bother to duplicate that action? Is it primarily to provide a more graceful failure? Or is the on-board 'DC leakage' going to be certified as safely protecting a type A RCD from saturation effects?