Mark is right, the JTR offerings would be great. They're essentially pro boxes, tweaked for HT.
Also, Bosso was correct as well. Generally pro subs have an -3db point around 35hz., give or take a little.
On the ultra cheap side, you can go Peavy. At less than $250 a piece, they offer a passive 18". (8) of these things and you've got a $2k sub rig you could power with an EP4000. (8) 8 ohm boxes, config'd 4ohm mono and powered by a mono Berry, would spread 250 watts easily across each coil. They are spec'd at 95db@1w@1m, so... 250w=119db@1m, .....and (8) boxes realizes a +18db advantage.... resulting in a 137db@1m, ....subtract 12db for a 4m LP=125db@LP@2kw.
Now all these are based on the factory spec of 95db. For a 35hz target, that spec may be reasonably close. When you add boundary gain and room gain down low, those figures would raise in level...potentially raise quite a bit. Boundary gain will get you back 6db per boundry,...all the way up to 18db fully corner loaded throughout the range. The driver needs to be within a quarter wavelength of the frequency involved to achieve full complementary loading. So at 100hz, the driver needs to be at least within 34" (1130/100/4) from the walls to achieve the full corner loading effect from 100hz. on down.
Any theoretical boundary gain one calculates must be lessened somewhat by the diaphragmatic losses of the structure. Which, I've got no good info on. So, this array aligned along the back wall perhaps, this would yield 12 db minus losses, which brings us back up to 137 db@ the LP
before the boundary losses. Not bad. Also, spreading the power across so many coils, retains the snap and responsiveness of low compression. This would be a fun exercise, with a $250 18" sub.
Also, the box would likely benefit from some additional bracing and perhaps stuffing. One could experiment.
My 2 cents
btw, those numbers may be off, that was one pass thru with a spell check only