Regulator first stages have an intermediate pressure of around 120 to 140 psi and this is normally regulated by spring pressure. They also incorporate an ambient pressure chamber to allow the increasing water pressure at depth to assist the spring and ensure the intermediate pressure ramains 120 to 140 psi above the ambient water pressure. This ensures regulators breath easily at depth.
The problem is that the expansion of the air as it goes from say 3000 psi to 140 psi requires a lot of heat and it draws this from the regulator and eventually from the water itself. In a first stage with a flow through piston design, the piston can get very cold, and when it lifts off the seat, a small chunk of ice can form behind it in the ambient pressure chamber and prevent it from returning. This then prevents the stem from seating and shutting off the high pressure air supply. The resulting excess intermediate pressure causes the downstream valved second second stage to freeflow to vent this pressure and the continued high airflow leads to more ice formation which assusres the process wiill not stop until the air is shut off or the tank is dry.
Traditionally, silicone was used to fill the ambient pressure chamber to allow the silicone (which won't freeze at the temps involved) to transfer the ambient pressure to the back side of the piston while keeping the water out of the ambient pressure chamber.
Current practice for example with Scubapro piston regulators is to use various plastic spacers and bushings to insulate the piston and spring from the cold with what is called their TIS, or thermal insulating system. A non petroleum based lubricant (christo-lube) is also spread on the parts to prevent ice formation. This approach solves some problems with petroleum based silicone that may not be compatible with higher O2 levels, but is not as effective as the older silicone filled SPEC system, assuming the silicone level in the SPEC system is properly maintained (some leaks out on every dive and needs to be topped off now and then.)
Other manufactures use a silicone liquid held in place by a rubber cover or diaphram in a piston first stage, or just use a diaphram to prevent water from contacting the critical parts in a balanced diaphram first stage.
In warm water the temperature differential is high enough to ensure adequate heat transfer from water to regulator to prevent the temp of the first stage parts from falling below freezing. But in colder water, or with very high breathing and/or flow rates, the required heat may not be transferrred fast enough from the surrounding water with the result that the first stage may experience some icing, and just a bit in the wrong place can cause a freeflow.
If you are diving in deep/cold water, it's a good idea to use a regulator designed for it. And even then you may need to watch your breathing rate and try not to do things like inhale, fill your BC or inflate your dry suit in one long burst of air as it can chill things excessively. Small bursts on the inflator and slow and easy breathing will go a long way toward preventing freeze ups.
I have also noticed some divers have more trouble with freeflow in cold water with nitrox. I have not dug out the physics notes yet but I suspect the differnece in specific heats may be a contributing cause of this.
Most modern regulator breathe well down to pretty impressive depths. Air gets "thicker" under pressure and does not flow quite as well but most modern high performance regs have smoother corners and larger air passages to accomodate this. The original recreational diving limit of 130 feet was based in part on the marginal performance of older regulators below this depth, but it really is not an issue anymore.