The crystal responds to the larger vapor gradients near the edges by sublimating faster there and thus rounding the crystal.
J. Nelson, J. Atmos. Sci., 1998, 55, 910-119.
The change in shape according to the gradient makes intuitive sense to me but I can't for the life of me figure out why it does this. Maybe I'm being really foolish here, but surely a higher concentration of vapour at the corners would decrease the rate of sublimation there.