Just a couple of thoughts in reply--I'm more interested in getting you
to go try the thing and let us know how it works for you! Though I see
problems with it and would personally go other directions, I can see it
could be quite a bit of fun to play with if you're interested in it.

So, the Q of a 1/4 wave coaxial resonator (and as it turns out, also
1/2 wave and 3/4 wave and...) built from smooth copper conductors,
paying attention to making very good electrical contact especially
where the current maxima are, and making the structure be about 77 ohms
(D/d about 3.6) should allow an unloaded Q of about 107*D*sqrt(f),
where D is outer conductor inside diameter in inches and f is the
frequency in MHz. So a 6" diameter copper tube with a "1.5-inch"
copper pipe center conductor, at 100MHz, should give you a Qu of 6000,
roughly. That would be a bandwidth (unloaded) of 1/60th of a MHz, or
about 16kHz. Maybe that will help a bit. In general, for moderate to
high Q, bandwidth to 3dB points is equal to center frequency divided by
Q. But that's loaded Q, which will be less than unloaded. Of course,
your regeneration raises the effective Q--it is after all a Q
multiplier.

Oh, and don't bother trying to silver-plate the resonator. You're
better off just making sure it's smooth, high conductivity copper
(probably NOT that hard copper water pipe I mentioned above!) with very
good contact everywhere, protected with a thin coating of clear acrylic
lacquer to keep it from tarnishing. There have been controlled
expreiments about silver plating that show it actually can make things
worse, unless you get all the plating parameters just right.

The regenerative cavity thing was/is a receiver for pulsed radar, where
the time of the return pulse is known, roughly, and tracked, so the
regeneration is gated on at the expected time of the return pulse.

Cheers,
Tom