Finally got a log recorded on this quad. Here are the results after running the log through Plasmatree PID analyzer.
One thing I notice is just how bad the lower rate pitch and yaw step responses look. I wonder if that is from not being so aggressive in the flight this log is from. I did 10 rolls, flips, and yaw spins this flight. Usually I just fly a pack like normal, with all sorts of speeds, tricks, and mixing of the sticks.
The filter transmission on the right of the noise plots is also very concerning. That should look more like the raw gyro noise, with most noise near 0 hz, and the rest quickly tapering to 0. Instead, you can see in all the noise plots that the noise is pretty constant, and higher than desirable, everywhere. I know this build, and KISS in general, has D gain oscillation problems, so I wonder if that is feeding back into the system. The D gain oscillations are very obvious in flight when I cut the throttle.
I am going to replace the soft mounts for my FC with the original hard nylon standoffs, as Psyborg suggested. That seems counter-intuitive to the noise I am seeing on these plots, but I'll give it a shot. I am also going to put on some Emax Avan Flow props and see what happens.