Full Duplex QRSS TRX

I was dreaming about this for a while (see earlier entry), get some inter-activity into QRSS. What about having full-duplex QSOs in a single band? I figure, given the narrow bandwith, this would just be possible.
All it needs is in-band crystals for filtering.
We have got a couple of in-band crystals available, here are some options:

• 3.500 & 3.579
• 7.000 & 7.159
• 14.000 & 14.318

Those pairs have a particular advantage, they are harmonics. Which allows for some neat designs. One could image multi-band rigs with those, e.g. direct-conversion on 80 and subharmonic (doubled) on 40, or the same with 40 and 20, or, have a direct-conversion tri-band rig with oscillators running on 20 while the frequencies could be divided by FlipFlops.

A possible DX-rig could be 20 and 40, while a NVIS-rig could be 40 and 80.

For single band TRXs there would be more options, such as

• 3.579 & 3.868
• 28.188 & 28.322

or any other widely available pair (check out the shops run by N4ESS or VK1AA).

To the design of such a system, the following could be said, one would either work with changeable modules, or build two full transceivers.
The latter option could be a simple a building two Rock-Mites. Those little transceivers could easily be modified for FSK. Another solution could be the PSK-Warbler design; unfortunately, the kit has been retired by now.
Modular could be somewhat like this: modules for the oscillators and filters and a common mother-board carrying the mixers and the rest. The easiest way would probably be to build one module only such that it can be mounted in two different orientations. That would just required the pins of the module being symmetrically arranged. In that way, one could easily change from LOW-TXing to HIGH-TXing by just reorienting the one module.
Another idea would be to build two full receivers and one switchable transmitter. With would allow to constantly monitor both frequencies, i.e. with individual spectra for left/right audio channels under Spectrumlab.

The remaining challenge would be to decide for the aerials. I figure reception would be done best using magnetic loops or other very narrow-band contraptions, while transmission is probably done best with a vertical of a dipole.

Anyone prepared to join the fun and build something alike?