I wonder if this is tied to the (alleged, highly dubious) effect where hot water freezes faster then cold water? I always figured if it was real, it had to be some momentum-based phenomenon. So if (a) hot water in a freezer sets up strong convection currents and (b) those currents are more useful for cooling the water than the "usual" reverse convection of cold water (by bringing warmer water to the surface where its less insulated, or at least not actively burying it) and (c) the regular convection is able to continue on momentum even when the water has cooled a lot, then it might conceivably accelerate cooling enough to win the race. Maybe in just enough of a patchwork, fragile way to match the mess of inconclusive tests about the whole phenomenon.