I don't know the current circumstances, but I've read a lot of older accounts of the high levels of RF on the 40m band in Europe due to the use of the high end for SW BC. That means that your front end and mixer need to be able to handle very strong signals.
I am managing OK on 40m at my southern England QTH with the little three-transistor regen, but that is on a long wire (the dipole, and the pole it will be hoisted on, are under construction). I am not getting any breakthrough from SW BC. The only overload I'm getting is with the stronger SSB ham signals, which I can cope with by using a custom attenuator that I built: a switchable antenna trimmer and potentiometer combination, wired across the antenna input. I'm not planning to operate phone in any case, and the sideband signals are far up the band and not affecting the CW portion at the bottom.
The little regenerative set is plenty sensitive, but there are two obvious limitations with such primitive technology: (1) the white noise from the regeneration, and (2) lack of selectivity on CW. The second of these two factors is the one I am most interested in improving, hence the focus on building or obtaining a good filter for the planned homebrew superhet. When I get on the air, I want to use the regen for the first burst of (arduous) QSOs but I will need better "ears" pretty quickly. The only caveat is all of my equipment must be homebrewed.
In terms of grinding my own crystals, I have a whole pile of several dozen old FT243s (and older form factors) but only one of them is anywhere near 3300Khz. I assume you would want crystals that are a bit lower in frequency, to grind them "up". The one I have is gigantic, a military format with a pull handle at the top, and I think it is 3272.5 or something like that. None of my other crystals is even within 1Mhz below 3300.
I will think about it, but it strikes me as fiddly work considering that I want two crystals only 500Hz apart. I don't have any test equipment to measure/generate oscillation to that degree of accuracy. Great fun, if you can get it to work!
Edited to add: I just checked the resurrected AF4K crystal online shop, and it's the same issue: a big gap in the available frequencies, jumping straight from 2MHz to 3.5MHz with nothing in between.
I remember some years ago reading an account by N2EY about building his hollow state transceiver. He found a whole bunch of crystal filters at a hamfest, and built the rig around their rated frequency, rather than the other way around. That is probably a saner way to go about it. There are lots and lots of filters available at 10.7Mhz but I assume they are usually used in double-supers. Also a fair number in the 8Mhz+ range, often as drop-ins for Elecraft rigs.
BTW I am not wedded to the "Miser's Dream" design. Just thinking aloud at this stage. Using a proven ARRL design has been my path most of the time until now, but it doesn't always make sense, especially when parts are unobtainium.
Edited to add: By adapting/designing my own receiver, using book learning and help from this forum, I would learn a lot more than by just blindly building an existing ARRL design exactly. I had a great experience some years back doing something similar, building a more elaborate regenerative set using subminiature (battery-powered) "pencil" tubes. That set will also see action in my shack soon, although it is currently still at the "breadboard" stage (huge and unwieldy). It performs better than the ARRL three-transistor Novice set.
73 de Martin, KB1WSY (for now)