I've spent some time looking at a lot of aspects about this and it's rather obvious now why sticking with the ax5043 is so attractive. Finding workable parts is really hard, if not impossible.
The processor/FPGA is pretty easy. Due to the automotive market, many options exist.
The ADCs and DACs are harder. There is the audio codec, the TI TAC512. Its bandwidth is 96KHz, which is less than I would like. And I don't know if it's going to handle DC offsets (I'm suspecting it won't, but I can't find any info in the datasheet); if so there will be a hole in the middle of the frequency range that's unusable. Maybe that's ok.
I have found an ADC, the Analog AD4680, and a DAC, the Analog AD5541A, that would also work. They will go to 1 MHz and will interface directly with the DSP. The AD4680 has a bipolar analog input, so it has to have negative voltages that have to be created, costing a couple of extra small chips to do that. For ADCs, that's about the only one I could find. Very few are temperature rated and most won't interface nicely with the TI DSPs. The slower speed (<10MS/s) ADCs are really designed for industrial control, the faster ones are workable but expensive and power hungry. For DACs, a few options are available, but not many. Again, mostly designed for industrial control.
The really nasty part is the RF side. There's just nothing I could find integrated available that is temperature rated to what we need and is suitable. As Jim pointed out, the CMX991 might have problems with availability. I'm also a bit worried that it's not going to work at the temperature range. I can think of three reasons that the chip wouldn't work full duplex: It would generate too much heat, there would be internal interference, or it would not be able to draw that much power. So I'm thinking it's not going to be feasible.
So for the RF front end, I found the Analog LTC5562 mixer which is rated to 105C. That was it. It might be possible to use it for I/Q conversion, too, or it's possible to use direct conversion. Direct conversion isn't so great for narrow bandwidth, but it might be usable. As far as I can tell, there is no other available mixer with an extended temperature range. I mean, not even a 612 mixer. Well, I did find the MC1496B, but it would probably only go to 10MHz.
So you might end up doing your own gilbert cell mixer. (For I/Q, you can't do ring mixers as they have to be transformer coupled and that doesn't work for I/Q since you need low frequency and DC, really). Maybe a mixer could be designed with discrete parts on its own circuit board and mounted vertically on the main board? I don't know. You start doing all this stuff and it starts taking up a lot of board space.
If you need frequency agility, then there are yet more issues with RF synthesis. But I was assuming the frequencies would be pretty fixed on the satellite, so you could just use standard fixed oscillators. Though I haven't looked much at that yet.
So I don't know how feasible this will be.
-corey