Preflight Fail: High Accelerometer Bias - Custom Rotation

We have developed a vehicle that has the CubeOrange slightly back and above the vehicle’s COG at a rotation of Pitch 135.

We added a Pitch 135 entry to the rotations list, as well as in our custom mavlink forks, and QGC forks.

All that seems to work well.

Calibration appears to work correctly.

However, after calibration, we get oscillating preflight failures from high accelerometer bias.

The high bias appears to be isolated to the Z-axis.

This happens with multiple different cubes.

I’m wondering if we could have missed something with the addition of our custom rotation that could be causing this?

Or, could this be the result of a not-perfect calibration? How precise does the calibration procedure need to be? It can be tough with larger vehicles like this to get exact angles.

Or if there are other ideas on what could be causing this?