Resin doesn't really shrink much when it cures in a mold anyway, at least, the casting resins I have used don't. Plaster and hydrostone are also rather dimensionally stable. If anything you would have problems with the resin foaming and expanding from contact with water left in the print. Unless these "fill-behind" chambers are very large with adequate escape vents, you also would have problems with air bubbles and getting your chambers to fill unless you also cast in a pressure chamber, so I would think the raft idea above is more feasible? Automating the creation of the fill behind chambers in the model might be challenging as well (i.e. like implementing the auto support features of various slicers).