Ken,
Have you tried using only one tuner? I only had one dongle plugged in when I set this up the other day. Yesterday, I thought to myself... two dongles should be better than one right? No.
I also thought to myself that I read somewhere in the manual (I had read it in this forum) to switch the channelization to "heterodyne" when you are experiencing processor weaknesses.
I spent several hours yesterday plugging and unplugging and switching channelization. I have two RTL-SRD's and one NooElec. Ultimately what I ended up with was: two dongles set to heterodyne. I set the program to prefer the RTL-SDR for control and the NooElec to center on the system's center. The net results was a only a slight increase in processor power over the NESDR alone in poly-phase.
When in poly-phase I also found that my NESDR SMArtXTR (with the e4000 chipset) dongle works better than my RTL-SDR dongle. It seems to have a wider range and the program spends less time moving the center frequency around.
As to launching an X session by ssh. I have tried in the past. That isn't possible except I looked it up and there are solutions. https://askubuntu.com/questions/633782/how-to-start-x-application-from-ssh . I don't need that capability and don't feel the need to go installing other X managers right now. I have had bad experiences having more than one desktop environment at the same time. That said: I also found that SRDTrunk doesn't seem to save whether you had the waterfall on. When I start a new X session and launch SDRTrunk it seems to come on with the waterfall. Can you pass instructions to SDRTrunk at launch? More testing is needed.
I see in your screenshot that you have the playlist editor open and the program scanning at the same time. I still don't have luck with that. If I want to edit the playlist I have to start the program from the terminal and before SDRTrunk starts scanning, I make my changes. Then I start the scanning. If I need to make other changes, I have to stop the program and start it over again. I get java "out of memory" errors, even though the RAM is only half occupied. I think it means that JAVA is addressing memory with the processor and the processor is pegged out.
I hadn't noticed the program lagging decoding that severely (20 seconds) but I hadn't left two computers scanning at the same time to compare. I only have one antenna.
I disagree with the Pi4 not being up to the task of being a full time scanner(except for the lag thing). I am quite happy with its performance. Now.... I have no need for a full time scanner so I might be biased. I too have a beaucoup computers sitting around here so needing the power to perform the work is not an issue. I do think that we will see more ARM computers in the near future with Apple switching over. With that, we will also see more development in the software. I would think that Sun is already elbow deep in ARM development.
Again, good luck and happy scanning,
Zach