Sorry I didn't reply earlier, but I did take a look and your magnetometer data is very wrong, no question there. The accelerometer and gyroscope seems to suggest that the device is at rest, however the magnetometer offet changes radically. This MUST be coming from external sources.
The magnetometer calibration status also reflects that. When it is 0/3, the magnetometer data in ignored in the sensor fusion, and heading drift is inevitable.
Hope that helps you find the problem.
We'll take this into account and be sure to eliminate any external sources when these issues come up. However, these two sensors/boards were placed in the exact same environment, side-by-side in fact, while the test was run so I'm not sure it is as black and white as it may appear.
Further digging required on our end..