Some people at UC Berkeley have been doing some interesting work with SDR that kind of mirrors what's been going on in the amateur SDR community: They have concluded that building custom systems takes too long and costs too much money, so they've started using the same kinds of components and techniques that the amateur HPSDR community is using. Check their web site at https://casper.berkeley.edu/
As an example of the kind of things they are doing, consider a 44 receiver array with beamforming and correlation centered at ~1400 MHz with a 200 MHz bandwidth! This takes a _lot_ of processing power, but the compared with what can be done with a single receiver, the contrast is dramatic. To me, this validates the approach of using high-speed direct ADCs and large FPGAs performing the DSP "heavy lifting". I also notice that they are using the Polyphase FFT filtering being championed by Frank Brickle and Robert McGwier for the AMSAT DttSP project.