johnhh - 9:44 pm on Nov 21, 2011 (gmt 0)

It appears that the Googlebot will follow any link, anything in text or an actual link, in our case we have "..." at the end, anchor text mixed in with the link, html code included and a bad case of a trailing / after filename.htm. Sometimes these actually resolve - which then create a load of file not founds as the relative paths are incorrect.

I would have thought they would have discounted these at once as they come from off-site, but no, the bot aimlessly follows them.