I wouldn't try to predict anything on that timescale.
The processes that eventually result in an oxide filament
or rupture take time to evolve.
HBM with its limiting resistance and large backing energy
(C*V) is very different than CDM (no limiting, small C) as
well.
I'd be looking for ways to get that pulse, or close to it,
and take the data. You're going to see that testing
eventually I expect, better to be on the right side of
it.
I try to get vgmax (ESD damage) on 1ns pulse. If I can get CDM (10n), (5ns) data, how to calculate the 1ns data?
Is T2/T1=(V2/V1)^-n the right equation to predict?