The freeze issue is not simply one of piston versus diaphragm. There are sealed and unsealed piston designs as well as sealed and unsealed diaphragm designs and freeeze resistance varies greatly even among sealed and unsealed designs.
Most, but not all, piston designs are unsealed, but some unsealed designs are extremely freeze resistant. The Scubapro Mk 2 is a good example of a very reliable yet unsealed piston design that does extremely well even in 33 degree water.
Many divers almost automatically assume that all diaphragm designs are sealed when in fact many diaphragm first stage designs are not sealed. Many diaphragm first stage designs also come in both sealed and unsealed versions. Aqualung first stages are a good example as many come in unsealed normal versions as well as sealed "Supreme" versions.
And like some piston designs, some unsealed diaphragm designs are very freeze resistant. The Scubapro Mk 16 is a good example of an excellent unsealed cold water diaphragm reg. At the same time some sealed designs may freeze very readily if the seal over the ambient chamber leaks and floods or if the alcohol or silicone oil inside leaks out.
So all in all, cold water freeze resistance really needs to be considered separately from piston/diaphragm design.
Generally piston designs tend to be comparatively simpler with few moving parts and generally less complicated design. Piston designs can be found in balanced models which are capable of very high performance as well as in very simple unbalanced models.
The advantage of an unbalanced piston designs like the Scubapro Mk 2 or Aqualung Calypso are extreme simplicity and reliability combined with minimal spaces in the regulator exposed to high pressure air. In these designs, there are only 2 dynamic (moving) o-rings and neither is exposed to pressure differences greater than 145 psi. The downsides are that these "flow by" designs have comparatively low flow rates around 80-90 SCFM and the intermediate pressure will decrease as tank pressure decreases leading to slightly higher inhalation efforts at low tank pressures.
High performance piston designs like the Scubapro Mk 25 have extremely high flow rates - around 300 SCFM in the case of the Mk 25. But they have slightly higher parts counts and one of the two dynamic o-rings in the balanced "flow through" piston design will be exposed to full tank pressure. This can create issues with high pressure o-ring "pinch" at high service pressures (over 232 bar / 3400 psi) and requires very close manufacturing tolerances to avoid the problem.
It's almost impossible to find an unbalanced diaphragm design as it is extremely simple to balance the seat carrier in a balanced design - so simple that it is almost impossible to justify not balancing it.
While piston designs use a down stream seat design where tank pressure helps open the valve, diaphragm designs use an upstream design where tank pressure helps keep the valve closed. This has theoretical implications should the mainspring in the first stage break as a piston design would posisbly continue flowing gas (most likely freeflowing it) while a diaphragm design would be more likely to stop providing any gas at all. But practically speaking, mainsprings in first stages are very heavy, very durable and breakage is all but totally unheard of.
Diaphragm designs tend to be slighlty more complex overall than balanced piston designs and tend to be a bit more difficult to service due to the higher parts count and the greater diversity in layout and design that is possible (compared to only two common types of piston first stage designs).
They also tend to fall between unbalanced piston and balanced piston designs in terms of performance with flow rates around 125-180 SCFM. However it needs to be stressed that anything over 110-125 SCFM is gravy anyway, even in deep air and technical mixed gas applications. So the 300 SCFM produced by a very high performance balanced piston design like the MK 25 is essentially overkill and is much like taking a Callaway Corvette to the grocery store.
Basically, if you buy a reg from a major company like Scubapro or Aqualung, it's pretty hard to go wrong regardless of the design.
Personally I am a big fan of the Scubapro Mk 17 and prefer it to the Mk 25 due to it's sealed design and exceptional cold water performance. The G250, X650, S555 or S600 second stages are all good choices for it.
I also have nice things to say about the Aqualung Legend first stage. It is well designed and has very good cold water performance. Aqualung's adjustable second stages are also well done and use an interesting poppet/shuttle valve design that seems to combine old and new technology to produce a second stage that performs nearly as well as a G250 or S600.
Since both the Mk 17 and Legend are diaphragm designs, I am obviously currently biased toward diaphragm first stage designs - which is very interesting considering that I have been a confirmed balanced piston first stage fan for the previous 20 years. I think that says a lot about the quality and performance of recent diaphragm designs as well as a lot about a general plateau and even regression in cold water reliability that seems to have occurred among current balanced piston designs.