Have anyone been able to determine (or guess on) the length of the "transitional period". I only had time to read Bill's excerpts so far (thank you for the find, Bill!) but it all sounds to me like this:
If you are making a change (any change) to a site, you're entering a so called transitional period, during which the result of the changes is randomized. [side note: for this bizarre patented technique to work, Google has to intentionally make the SERPs worse in at least 50% of the cases of page changes]
So, they hope that if you spam, and if the randomizing worked NOT in your favor, then you have to flinch during the transitional period and remove the changes, thus they'll be able to mark you as a spammer.
How does that help them if the spammer spams "long term" (for the lack of a better term) - just creates hundreds of automatically generated sites that are not going to be changed, no matter which way the random ranking went. So, he won't flinch (because he did not plan on making changes to the site anyhow) during the transitional ranking period, and 50% of his sites will rank, 50% will tank. Not a bad ratio if you ask me.
It seems rather strange that the long term rank will depend on the actions (or lack thereof) of a third party - not really what you'd think linguistics and computer science PhDs would come up with.