Thanks for the explanation, NevadaMark. Unfortunately, the system is at our customers and we don't have long-term access to it to measure the drifts. Do I understand correctly, that if we correct the bias and scale factors, the sensor's performance would restore to its initial values?
Also, you've mentioned temperature change as one of the factors. Could transporting the system in a baggage room in a plane be considered as undesirable? Another point is that the sensor heats up itself quite rapidly (maybe, not 15 deg/min, but still...). Does this affect parameter drift?
And the last thing, concerning mounting. Our system does not fullfil all your recommendations, and the biggest problem, as I see, is that we tried to organize some passive cooling to the sensor, so it is attached to a PCB on one side and through a soft heat-conducting material to a radiator on the other side of the sensor (like a sandwich). Is this really bad? Any recomendations about heat dissipation?