Need help for crash analysis

Hello forum, I recently had a really bad crash that has left me wondering what went wrong.

The plane was a hobbyking Go Discover, with a PX4 running ArduPlane V3.3.0.

It was my first time flying a flying wing, so I followed the wiki instructions on how to set up elevons. They moved fine in MANUAL and deflected as expected when ground testing in FBWA and STABILIZE.

I went to the field and did an OK hand launch in MANUAL, followed by a lazy turning climb at full throttle to about 60m. I then flicked the switch on my Tx to put it in FBWA. The plane pitched up aggressively, then pitched almost straight down and planted itself at full throttle. I fumbled with my Tx to put it back in MANUAL and pull up but I had taken my finger off the switch and didn’t manage to get it back in time, you can see in the logs it switches back to MANUAL at the very last minute before crashing.

One thing that might or might not be of importance is that I had misgivings about this PX4 board. I had previously installed it in a Bixler but when testing by rapidly moving the sticks all over the place, it would still move the control surfaces aggressively for a while after I had stopped. I don’t remember what mode I was in but I remember that I failed when trying to reproduce this behaviour on a Pixhawk. I tried this again on the Go Discover but it seemed to behave itself just fine. No idea if this might have anything to do with the crash.

I attach .log, .tlog and .BIN files. The plane was sitting on a bench for 97% of the log so the action is only at the very last bit. I can see that the desired pitch deviates wildly from the actual pitch. I wonder if the crash was the result of a stall, caused by the sudden pitch up. From what I saw, I would have expected a small stall and swift recovery but not followed by a 60m vertical plummet.

I don’t know what to look for in order to figure out:

1 - what the control surfaces were doing once I switched to FBWA.
2 - why they were doing what they were doing.
3 - what my inputs actually were (my guesstimate is pitch down when it was pointing up, followed by nothing (thinking FBWA would sort it out once the nose was down), followed by pitch up. However, I can’t really tell by looking at ch2_in in the telemetry log and I can’t find RC inputs in the dataflash.
4 - how can I tell from the logs how RC input gets translated to servo command?
5 - why don’t my dataflash logs include RC inputs? Are they not on by default? Do I need to enable them?

Any help will be greatly appreciated.