MQ Hydrogen sensor randomly drifts?

I have a hydrogen sensor, similar, or same as this one:

http://playground.arduino.cc/Main/MQGasSensors

Mine has a potentialmeter already built in. I have it setup fairly sensitive. What I'm trying to figure out is, the value is normally in the 30-40 range but at random it will start to hover in the 50-60 range for a few hours, then go back. I have a battery bank of 4 12v marine batteries inside an enclosed cabinet and the exhaust is directed at the sensor. The batteries are on constant float charge and there is no equalize schedule or anything like that. I have set large fans and that does not help.

I'm starting to think it may be the sensor itself acting up. Could this be? What could influence the reading of the sensor randomly like that? Are these temperature sensitive? In winter my server room tends to be around 10-15c while in summer it will be around 20-22c. The fluctuations don't seem to corrolate with temperature though.

Are these generally good enough for safety applications, if not, what would you recommend, that can interface with arduino?

Firstly until you have some independent way of measuring H2 concentration
you have no reason to suppose the sensor is misbehaving. Pb batteries on trickle
charge generate bubbles of H2 and O2 and these are going to rise to the surface
of the cell in a random way perhaps?

Place the sensor somewhere else and test again - if the fluctuations vanish then its
probably a genuine H2 signal.

Have you tried more than one sensor (perhaps you have a poor one?)

Have you read about "burn-in" of such sensors to improve their repeatability?

I think an important test would be to expose the sensor to a known concentration
of H2 and verify it generates a definite signal you can rely on - perhaps the
small fluctuations are too small to matter and can safely be ignored?