Danko, yes the simulation describes the issue and you know exactly what this is all about.It is possible variant of your problem solution:
<removed>
In case you are still active and interested, I was able to eliminate the phenomenon by using double frequency compensation. I did quite a lot research and testing on this (I was unable to find much existing theory about it). The unwanted capacitance in a solution correlates pretty well to the frequency and the rest is getting the math right.
So, in case you are still experimenting I wanted to let you know.