triggerfinger - 6:42 pm on Sep 17, 2012 (gmt 0)
Another clue: All the URLs seem to have vanity tld URLs redirecting to them. Is G trying to access the robots.txt of these URLs and instead requesting it from the deep page? Seems like a rather dumb idea for such a smart algorithm.
examplevanityurl.com -> 301 -> example.com/deepURL.html
examplevanityurl.com/robots.txt -> example.com/deepURL.html/robots.txt
Seems silly, no?