With reference to Bookworm's last statement, surely it doesn't matter that the clear air being flown through is unsaturated. It still contains water, however little.
So if it is -25 deg and applying carb heat raises the temperature to -5 deg, then the carb itself will be at about the right temp for water in the air (however little) to freeze on contact with parts in the carb and slowly accumulate. Again if this was the case, then it would be hard to melt the ice (without descending) , as the carb heat is already on full.
Now am I thinking right, or does the air have to be saturated, ie water droplets be present for the ice to actually buiild up?
This thread certainly makes you think!