As you've discovered, OS uses resistive humidity sensors. The resistance of the sensor is measured at a low frequency (typically somewhere between a few hundred Hz to a few kHz). The resistance must then be converted to RH. The resistance is a strong function of BOTH temperature AND humidity. This means you must have a stable, repeatable measurement of temperature at the same time. The little cpu in the sensor has some internal tables or a mathematical function that converts a combination of resistance/temperature values into RH.
If you want to see what this 2-dimensional calibration data looks like, have a look at this data sheet for a typical sensor:
http://www.rhopointcomponents.com/images/SYH-2R.pdfYou can also draw some conclusions about the shape of the errors you'll get if the calibration is off here. You can also see that a simple offset or an offset versus RH reading may not be enough. To get a full calibration you could swear by, you would also need measurements at different temperatures. Quite a daunting task.
A given part-number sensor from a given manufacturer will generally have a unique set of calibration curves so without knowing the part number/manufacturer used in OS sensors, it would be a bit of luck if you found a perfect replacement -- not impossible but not easy either.
Notice that the resistance becomes very large at low humidity values (10's of megohms). This makes these sensors a poor choice for measuring low humidity. Capacitive sensors on the other hand work well and low humidity and are not as accurate at high RH.
Although the hair gel trick may restore a sensor to the point it appears to work again, without a full calibration over RH and temperature who knows how accurate the result is? It might be very good or it might not -- you just don't know w/o a lot of measurements at different temperature & RH values. For many folks that is good enough however and so this could be considered a valid "repair". The other mitigating factor is that the original sensor is not specified to have very good accuracy to begin with, so a repair with so-so accuracy results may be as good as the original. Again, I'm not saying this trick is a bad idea -- just that it would take a fair bit of work to verify the results over a wide range of temp/RH values.