I have zero ideas about how this would work with a cryogenic mobile phase. Water would seem completely incompatible at first glance as you are pointing out.
This is probably a case where you just have to do it to find out what problems are gonna present themselves. I have been a part of multiple engineering projects from concept to customer delivery of some very sophisticated robotic motion control stuff for factories and mills and such. I learned with certainty that on a new project there is simply no way to anticipate the problems you are going to need to solve. In that case you just make your best guess and go for it lolz.
At times things go boom, or bridges collapse, or the Space Shuttle does blow up, but generally that is the way engineering is done. You make your best calculations, best judgements, and best effort then hope the boss is not present when the purple smoke starts coming out of a six figure machine slated for delivery in one week and already six weeks late…
My delayed thinking suggests that if you pursue this as you reminded me water may well be a huge factor for you. Therefor my input into the project would be only use activated alumina which is what I destroy by adding the water. I forget how hot they have to cook alumina to get it REALLY dry but my oven my home goes to 550F and as I recall the figure was above that at least a hundred degrees or so. Kiln temps.
I have found that even the slightest traces of an unwanted solvent say in chromatography can throw everything off to the point it is a fail. Tenths of parts of one percent by volume imo will change the way it plays out. If that notion carries over to your project then tiny traces of water in unactivated alumina would also impact it. Just a hunch here but that is why they go to the trouble of baking it lolz although I have only ever heard of activated alumina being a necessity in liquid chromatography as I might use it. I suppose you are building an LC project but one at radically different conditions than ambient so … liquid is liquid no matter the atoms but that suggests that any other liquid you use needs also to be only liquid at those conditions. The trick then would be to find solvents with a wide polarity difference but that both dissolve the THC and do so one better than the other.
Then you would have a mobile phase you can tune in like manner to DCVC. I better vape one more bag though before I edit more… my head hurts 
.