First off, imho, the way to do this is in situ. Meaning, in the actual power supply.
There are things that the SPICE simulation will not take into account. The dynamic
resistance of the power transformer, and the actual load Z and phase angle.
Also, elliptic filters, which is what this is are, are somewhat hairy in that the depth of the
notch (depth of the null) and the frequency is reasonably critical WRT to the actual
inductance and the actual capacitance. So, it's not really predicated by the simulation.
Think ur going to need a HV cap, depending on the required value, oil filled cans have
been used historically. Since it's a low frequency application, how they perform at RF
isn't terribly relevant, it would seem. A bunch of series'd film caps might fly?
I'd check the mfrs specs for various caps, some are AC rated, if not for a specific
cap, then for the type - say metalized polyester for example.
Some time back I found some large value surplus motor start caps that were actually
polypropylene, they'll handle large ripple currents at 60 and 120Hz... not useful for this
but just an example.
Oh, unless I am upside down on this, if the L decreases, you need more C to keep the
frequency constant.
If you look at the curve, you'll see that after the notch the attenuation more or less is
a flat amount, maybe 20db... before the notch the attenuation is reduced to zero.
As far as instantaneous current draw? Maybe a fast cap after the choke?
Btw, neat idea to simulate the QSK with the circuit you put in there!!