Bad IMU on Pixhawk? IMU v IMU2 discrepancy, drift

I’ve just got a new X8 Octo in the air, no drama on its maiden flight but the Log Analyzer reported this:

IMU Mismatch = FAIL - Check vibration or accelerometer calibration. (Mismatch: 5.53, WARN: 0.75, FAIL: 1.50)

I have very low vibes on this frame so I went through a full accelerometer re-calibration, re-mounted (with foam tape) and flew again. I did a very short test flight and checked the log, no error reported, so I flew again, more aggressively.

After flying around a while I brought it back to hover and found I needed to hold about 1/3rd forward stick for it to stay level (perfectly calm early morning conditions). Landing with the stick jammed forward I noticed that Mission Planner showed the horizon being way off (nose down) despite it clearly sitting perfectly level on the ground.

Here’s a section of the log showing pitch, you can see DesPitch matches commanded Pitch pretty well, and it is pretty level during the hover phase of the flight but slowly increasing during forward flight. When it lands it thinks it is pitching down about 15 degrees, though it was perfectly level.

When I disarmed I saw the HUD horizon quickly come back to level again.

Here’s the log from the initial maiden flight which reported the high IMU/IMU2 discrepancy:
dropbox.com/s/p6jkfvkykstw0 … 0.log?dl=0

And here’s the log from today’s flight which had the big pitch drift:
dropbox.com/s/7e5jvfpjudu0s … 4.log?dl=0

I saw a similar problem once before when I somehow had AHRS_RP_P set to 0, but I’ve double-checked my parameters and calibration so I’m at a bit of a loss. Could it be a bad unit? It has flown fine previously in a plane, taken a couple of hard landings but nothing major. One thing to note this morning was that is was about 5 degrees celsius, so could temperature drift be an issue as it warmed up?

I’m very nervous about putting this big ship back in the air when the flight controller seems confused about what level looks like! Any help appreciated, thanks.

I think I had a similar issue on my quad yesterday that ended up in a hard crash. I see some inconsistencies in GyrX between IMU and IMU2 right before the drift started. Did you come to any conclusion in your case? Thank you.

Cyril