Nawoj,
Servos receive their commands by measuring a pulse width -- how long a signal is high. The longer the pulse, the farther it turns the arm. problem is how do you measure time? if they used electrical parts insensitive to temperature, like a quartz crystal oscillator, it would have a very stable time measurement to know when the pulse started and stopped.
For the cheap blue servo, it's using a simple timing circuit -- probably not much more than a resistor-capacitor oscillator -- that is cheap but won't keep time well. since the clock runs slow in extreme cold, it thinks that medium duration pulse is now a long pulse, and your servo starts wandering.
The better servo likely has a nicer timing circuit that can ignore the cold and keep good time, so it stays rock solid.
moral of the story? watch out for cheap servos in extreme temperature changes -- better yet, leave the cheap servos out of the snow plane.