Ask HN: Redesigning set of examples for an open source project

Hello,

I am looking at redesigning examples that ship with a popular (large) open source project (Specifically, Apache Solr search engine).

I am looking for suggestions of what people would see as a best practice for an example set to lower the initial learning curve.

Obviously, the first example would be the most basic end-to-end run with smallest number of steps. But what else? Should it be an example per feature? An example showing complex dataset processing? A kitchen sink with a lot of documentation (already have those!)? A project big enough for somebody to test their knowledge against?