If you are asking why does C422 not go to ground, but rather to the output of the transistor, my own little research suggests that this is a bit more than a simple RC filter. It looks to me like the topology of a Sellen-Key filter like this: https://en.wikipedia.org/wiki/Sallen–Key_topology
Do I have it right? I can't explain exactly how it works though. Need to study this.
Quite right, it doesn't have the unity gain of an op-amp so the stop-band attenuation won't be perfect, but the transition band can be peaked for whatever filter characteristic you need, and the asymptotic slope can be -40dB/dec (until leakage takes over).
Note the feed-forward path through the first resistor and capacitor, so that at frequencies where the transistor gain is dropping off, this path dominates, limiting stop-band attenuation. Admittedly, this is still a higher roll-off than for a typical op-amp, so it might not be too bad.
Here's an example I made, with vacuum tubes for audio cred:
The 6V6 model is actually pretty close (when the model's bugs have been avoided..) to what was actually used, a 5702 submini pentode. The simulated response:
You can see it's much sharper and tighter than anything you can do with resistors and capacitors alone, and smaller than most things you'd do with inductors (I know, you can get pretty small ~1H chokes, though the Q is pretty bad). Plus you'd still need the buffer to drive an LC filter, so might as well do it in one.
I don't have a picture of the actual filter, but it went on
this radio, just after
the detector. The construction is similar, just imagine it with another submini and more capacitors and resistors around.
Also, incidentally, I finished up that radio by powering it with two SMPSs, in a hybrid configuration. One AC to 12V 2A supply (Meanwell) just to get things going, then a discrete 12 to 6.3 and 100V supply (a slight variant of
this) for the tubes. I added shielding around, and an extra stage of LC filtering outside the discrete module, and that was all that was needed to keep the thing dead quiet on the 20m (~14MHz) band.
Tim