A commenter on [...] chicagochess.blogspot.com may have found a blockade draw after all: 1.Nd7+! Kf5 2.d4 [...]
Yes, this was noted by several people on chessbase too, see the last paragraph of http://www.chessbase.com/puzzle/christmas2009/chr09-03.htm. I've seen at least one case before where a study contains both a refutation of the intended solution and a dual solution, so formally they cancel each other out and the study is sound, but rarely of nearly as much interest as the intention. Here the new fortress of some interest (in that the g-file comes into play as well), but we lose the spectacular self-stalemating point 4 Rh1! ... 7 Kg1!, and the move order in the introductory play is still not unique (2 Nf7 works too).
Perhaps Nunn should have found this, since he was tasked with vetting soundness and so might have looked for alternative solutions once he agreed that the intended solution works. For myself I might claim the excuse that I recognized the intended solution and, having found that it seemed unsound, stopped analyzing for further flaws...
--Noam D. Elkies