I've been wanting to restructure my category tags for sometime now, but haven't the time nor the organizational plan to do it. The category tags are the best and worst part of CiteULike; right now, there are so many tags that when I post an article I can't select all the relevant ones. On the other hand, as my knowledge of various topic areas has grown (partially as a result of my incessant literature searches, no doubt), I have found a need to reorganize tags according to how they fit together with other tags.
One example of this is evolution of development. I started a category called "homeotic-theory", which was supposed to be all articles on Hox-related topics (basically, the regulation of phenotypic development). After taking a class and writing a paper on the topic, I found this topic to be vague and no longer intuitive. However, simply going through every article and changing out the tag proved to be impractical, since there were already a significant number of articles (e.g. over 15) in the bin.
One particularly good feature of CiteULike is the ability to bridge disciplines. I can post articles from a number of fields in the same library, and once I have a better scheme worked out, I may be able to use my CiteULike library as sort of a "research planning engine", where categories and experimental approaches form building blocks which can be assembled to solve novel problems. I would like to teach a seminar on this (research mining) at some point. But the idea isn't well-developed enough as of yet.