The Wireless Log tab consistently shows "Noise" as -90/-91dB. I assume that's measured at the main router, but as I say the house isn't very big so it seems unlikely to vary much. That's good, I think?
That same page shows one connection at the end of the backhaul section, which I assume is the one from the "A" node back to the main router. It shows an RSSI of -63dB right now, and it's consistently between -60 and -63. I think this is OK but not great; it's also the best I'm going to get in this situation. I really can't run a wired backhaul; I've tried powerline and the result is terrible (the mesh protocol does *not* handle dropped packets well). I'm not going to spend $200+ on a *fourth* router for this tiny house, just to wire yet another node in the stairwell going up to the second floor or something, which is the only way I could avoid having the signal travel through the (extremely RF-hostile) floors.
I've run iperf3 for a pretty long time and gotten consistently good results, when the mesh network is "happy". I wouldn't even be mad if I occasionally got 10/20/30 second outages when some RF spike disrupted comms, and the nodes had to redo some kind of handshake. What bothers me is, nodes will fall off for multiple minutes at a time, then work perfectly within 90 seconds or so if I just power cycle them manually. I just want them to recover gracefully without manual intervention.