btw I'm pretty much sure I know how they detect such networks.
For example, we have a small-to-mid-sized network. We buy a big package of links on this network. This package represents, say, 1% of sites in the network. Maybe even less.
So, we flag 1% of sites as link sellers.
Then we scrap all outbound links on these sites. And look for other sites with outbound links which somewhat intersect by domain with this 1% of links on already flagged sites. 100% intersection is not needed. I think intersection with just a few percent is enough. These other sites are link sellers too and they are added to the pool of flagged sites. Rinse, repeat, several iterations.
Now we got 90% of sites in the network. They are "bad". Let's select all the sites which have they PR mostly from the bad sites and apply a penalty on them. Negative-seoed authority sites will mostly be not affected as they have significant part of their PR flown from other, "good" sites.
This process is cheap and requires only a small deposit to flag the first 1%. Rest can be automated.
This is why the target small networks so successfully. This is why SAPE (huge pool of very differentdonor sites, hard to find intersection) is not affected as much as small networks yet. This is why SAPE sites which sell more than 1 link/page and/or have many pages on SAPE are affected the most and the most dangerous.