trakkerguy - 11:33 pm on Jul 13, 2011 (gmt 0)
I don't think the results imply that the "excellent" authors recovered because articles on similar topic were grouped together.
It seems much more likely that they were judged as quality articles, and no longer suffered from being seen as related to the "weaker" authors that were most likely seen as crap by google.
Say you have 20 excellent quality articles, but they are on a domain that is 60% crap. Panda (unfairly) strikes them along with the crap.
Put them on a brand new subdomain. They get a fresh start and Google evaluates them on their own merits. All good, no Panda.