G-Man has an interesting post about supplemental results and PR. From personal experience, I believe there are a number of factors that contribute to pages ending up in supplemental hell. Two in particular are low page quality-score and ineffective site-linking structure.

Page Quality Score
Not too long ago, I had tons of pages on many, many splogs go supplemental and the overall traffic levels dropped significantly. After some testing, I found the problem was down to using singled-chained Markov content. I experimented with mixing legible and Markov paragraphs, and discovered that the more single-chained Markov content on a page the greater the chances of it going supplemental. This might have ramifications for poorly written content.

Site Linking Structure
Some of my old WH blogs have hundreds of pages each, and a fair number were supplemental. All I did was simply reduce click distance, by linking to individual pages from the sidebar and lo and behold, many of the supplementals went into the main index! This demonstrated that improving site linking-structure and distributing link-juice certainly helps.

It wouldn’t surprise me that the recent, aggressive nature of the supplemental index may be a result of Google’s attempt to reduce spam in its main index. Also, Google is probably using the supplemental index to inform webmasters of potential problems with specific pages, including low quality-score, poor linking structure and duplicate content issues.

These points might be worth considering if you have many pages in supplemental hell.