"You assume that the process of RMNS can program new and more complex creatures at a staggering rate of success. We are talking about millions of highly sophisticated functionally complicated systems."
This is not an ungrounded assumption given the tried and tested ability of RM + NS to design, and the millions of years of time available for it to have done it in. On the otherhand assuming that it is impossible before understanding the full picture is a bad assumption to make.
I discount all the examples listed except the goatsherd, which may very well fit the prediction: A new speciation within an existing population at a given moment in time. Unfortunately, Talk Origins sometimes plays loose with the facts. I'll have to see if I can dig up the original Scientific American article. Similar cases in the past have turned out to be nothing more than discoveries of an established species, either by stumbling across it, or by changing the definition for what constitutes a species within the genus.
If we can find one example, that strengthens the theory. The next step would be to determine if the percentage rate observed matches that predicted in the fossil record.