[morzh]

Oh.....RF engineers..... No, no, they are great folk, and myself I do not know the first thing about RF, this is a black magic for me, but as I remember those I worked with, power supplies for them is as much of a black magic

In fact when each diode conducts the respective capacitor connects in parallel to the last filter "reservoir" cap (40uF) WHILE the first filter cap (16uF) is being charged. However 3nF to 40uF is 1/13,000 times difference and this is how much extra "reservoir" action you get.

However the RF interference that gets through the capacitive barrier between Primary and the Secondary, usually as a Common mode noise, will appear on every single lug of the transformer's secondary windings (and will be approximately the same).

the main reservoir cap (16uF) will not filter it out as it is common mode.
And the Ground (Chassis) is decoupled from the transformer (from the centerpoint) by the Field coil acting as a filter, so the second filter cap (40uF) will see it now as differential, BUT as an electrolytic it is too inductive to filter it well, so it will appear on the B+ relative to Ground.

This is where the common mode filter comes in, these two small caps will short the Ground and the transformer at RF frequencies, effectively eliminating the common mode becoming a diff. mode at B+ relative to Ground due to the Field Coil.

Now usually these caps are placed in the primary circuit, tying both L and N wires to the Chassis through 3-15nF caps. This however requires Y-safety caps (unknown at the time, there were no Y or X rated caps, hence the Death capacitors). So maybe this was an ingenious way of defeating Death cap problem by placing all caps at the secondary side.

Of course the rating of these caps had to be much higher, as each half can easily provide 350VAC when unloaded.