I've been reading results of temperature tests looking at variability of various powders so thought I would give it a try.

I have been working up a load for a new Christensen Arms Mesa in 6.5PRC. Here it is:
New ADG brass, Fed 210 Match primer, 57 grains of Retumbo, 2.95 COAL, with Hornady 147eldx is giving me 2806 fps.
So, I put three in the freezer at -20 for 3 hours, packed them in ice and went to the bench (25 feet away) and sent 1,2,3 as fast a possible before warming them up. First shot was 2811, second was 2858 and third was 2870 for an average of 2846 or 40 fps. faster than a 70 degree group with the same load.
FPS should have gone down. Any thoughts???

Slower powder burn, charge fully utilized over length of barrel. Rifle length gas? I think that you would see a drop in Velocity with carbine/mid gas.
SWAG

Maybe case capacity changed a bit do to cold contraction?

Retumbo is one of the Hodgdon "Extreme" powders (link).  The tables located HERE show some MV and pressure data for some of the Hodgdon Extreme powders compared to commonly used powders from other manufacturers.

Your observations are not uncommon (qualitatively) among "extreme" powders although I can't explain why...

Mine was certainly slower at cold but with W748 so....
8 degrees vs. 120
Attached File

.
.
Baseline temp 83
Attached File

WST pistol powder is known to be reverse temperature sensitive

but i like it anyway

I was talking to a hogdon tech a few years ago, and he told me how they accidentally discovered that VARGET was very temperature stable

typically IMR stick type powders are more temp stable than ball powders like BLC2, win748 etc

Nice first post, I always thought powder was more temp sensitive than this to the point i'd keep them warm next to my body or with hot hands, guess i can stop doing that.

Depends on the powder. You should really test first. I've seen powders that do this kind of stuff in one load but not in another that's otherwise pretty similar. When they start tinkering with sensitizers and desensitizers it throws the whole notion of predictability into the dumpster.