I did a little digging, and this question is not easy to answer. There is much material regarding atmospheric drag on satellites, but those calculations require the assumption of hypervelocity. I wonder whether a simple calculation based on particle density and momentum transfer, with the balloon essentially sweeping out a vertical column of atoms as it fell, would yield a good order-of-magnitude estimate on the drag force available.