Here's my take on it, which may be completely wrong, so feel free to shoot down in flames! Here are pentode characteristics (an EF86) showing anode current versus control grid voltage, for different values of screen grid voltage. Your NSC control allows you to vary Vg2 between normal value and zero. From the characteristics, you can see that reducing Vg2 reduces the possible swing of Vg1 before clipping occurs. I think the idea may be to have some kind of variable-threshold limiter, since most of the troublesome noise would be spiky amplitude variations such as ignition noise. This might be a way of reducing the impact of that. For some reason it never caught on...
Ive just looked up the 57 datasheet. The 57 Pentode is a sharp cut off valve, therefore when tuning between stations, the first 57 tube biases the 2nd 57 to cutoff, therefore muting the audio. I'm guessing the NSC control adjusts the threshold of the muting by adjusting the gain of the first 57 through the screen grid voltage ?
The first 57 valve, 57(1) is a quieting system which suppresses the output of 57(2) when tuning between stations. With no signal, the anode of 57(1) draws a small current from the cathode of 57(2) via the 500K anode resistor, so anode voltage of 57(1) is negative w.r.t. 57(2) cathode, and this voltage is applied to the suppressor grid (G3) of 57(2) which decreases the output of 57(2). With a signal, G1 and G3 of 57(1) become more negative, because the top diode of 2B7 is drawing current through the 500K diode load. This decreases current through 57(1) so less current is drawn through the 500K anode resistor, and hence the voltage applied to G3 of 57(2) is less negative. The overall result is to decrease the output of 57(2) between stations.
Just released part 3 and had my first copyright strike from YouTube - 1920's music I got from FreeMusicArchive which were listed as Creative Commons licenced and free for use, which apparently are not. Sigh... so much for making the videos more interesting. I've taken it down and will probably have to use some bland modern music instead... unless anyone has a big band or orchestra and wants to do some free music
There are no personal problems that can't be solved with the liberal application of high explosives
I've test-3D-printed a couple of hollow formers that take an orange-drop (the style of cap I use) with slightly smaller dimensions than an original RCNZ capacitor, to allow for the paper thickness. I need to come up with a repeatable way to roll the paper at the ends or maybe I'll just let the wax make it look near enough...
I'm thinking longer leads, a bit of hot-melt glue inside, print the label a bit darker on newsprint-coloured paper... dip it in wax... done.