Thanks for the link to the paper with the percolation model. I think it's interesting, but the assumption of independent probabilities at each stage seems relatively implausible. You just need one civilization to hit upon a goal-preserving method of colonisation and it seems the probability should stick high.