I agree that allowing independent invention to be a defense to patent infringement would solve almost all of the problems we see with software patents. But it seems to have some different trouble.

If you allow competitors to use the invention if they pay the cost of duplicating the research, you cap the value of the patent at the cost of doing the research rather than the value of the result in the market, because anything more than that and you cause competitors to [band together to] duplicate the research rather than paying you for a license. In general this is a good thing because monopoly rents are bad, and it allows the original party to recover their costs by licensing without allowing inefficient monopoly rents in excess of the original cost. The trouble seems to be that it doesn't provide an incentive to actually be the one who does the original research, because if the research pans out then you can only recover up to the cost of duplicating the research, but meanwhile you have to take the risk of the research not finding a solution.

The other trouble is, how do you prove or disprove independent invention after something is published?