After a few minutes of ground tests, my results seem to confirm your findings that the APM2-ch6 is involved in the ESC speed glitches. So I'm very glad that you shared this discovery with us.
Here's my set up, for the sake of completeness:
I use Spektrum DX7 transmitter with OrangeRx R910 receiver, with the first 4 channels connected to the APM2 in the usual way.
The RX's ch5 is connected to APM2's ch6. (I don't actually use this feature, but it is connected, this channel is controlled with the GEAR 2-pos switch of my DX7, according to the config page of APM Planner, it toggles between values 1922 and 1120 ms)
The RX's ch6 is connected to the APM2's ch5. (I use this for flightmodes, 6 flight modes following the instructions posted on these forums, using a 3-pos switch and a 2-pos switch.)
The RX's ch7 is not used.
TLOG & RLOG are attached. I use small cyclic-stick movements to induce the glitches. And I switch between APM2-CH6 high and low every 5~10 seconds or so, showing that the glitches occur only when APM2-CH6 is given a high signal.
For me, the random ESC blips (very brief duration pulses) clearly do stop occurring when I give CH6 a low value, and they are easily reproducible when I flip a 2-pos switch on my TX to give APM2-CH6 a high value.
However, I now have a new problem. It is a random, less frequent, more sustained uncommanded pitch/roll, which caused two crashes so far.
The very quick glitches that I previously had never caused me to crash because the quad always quickly returns to level. But the new misbehaviour is more dangerous. It is like a loss of control. I will need more tests to trace the cause... I can't be sure if the APM2 is to be blamed for this.
After getting similar symptoms when I replaced the APM2 with KK2 and also during a bench-test without any FC (just a 1-to-4 servo wiring), I am convinced that my problem isn't the APM2.
I suspect it is ESC/motor-overheating.
Bench-test also revealed resonance vibrations with my sloppily assembled US$40 quad frame from OMGFLY.