We maintain an SDE for office-wide use with several hundred layers. Our preferred method of organizing the data is to separate it by topic (boundaries, hydro, etc.) so users have an easy time searching for what they need. The current method is to have the data broken up into feature data sets. This has been problematic since the beginning. Performance is slow, the whole data set gets locked from editing if any single layer is in use, and you can't store tables. Previously to that the data was organized into different SQL databases, one per topic. I don't know the technical details of that system but I know it was cumbersome to navigate and IT had some additional concerns with it.
Is there a better way to do this? We are getting ready for a server upgrade and it would be a good time to revisit the issue.