While debugging an issue on Pixhawk-derived flight controller our team noticed that zero-g temperature sensitivity of LSM303D is an order of magnitude higher than specified in datasheet (measured 5 mg/degC instead of 0.5 mg/degC as specified in datasheet).
Has anyone else came across this? Googling only reveals this thread (PreArm: Accels Inconsistent - #2 by Rob_Lefebvre - Copter 3.2 - ArduPilot Discourse).