When I want to learn something new, I often turn to PhD theses. Not all theses are equal, but some are very good to start with.
Very often, the introduction is quite naïve, in comparison to what a professor would write. Yet, I don't skip it. Every good thesis has the canonical references in the field cited there.
But, the most important part are the tools (experimental, and theoretical) that the student used for their work. In some theses, I can find a lot of details on how the results were obtained, and those details are often unavailable from other sources. There are also worked out examples, and results I could reproduce, that are described in detail in a PhD thesis.
As others mentioned, the best thing about a good thesis is that is accessible to people from somewhat different background than the author. In other words, a postdoc in Experimental Condensed Matter should be able to kick start his research in a field of Condensed Matter Theory, using a good thesis as a guide. With some more serious additional work and sources, a persistent postdoc should be able to learn a lot from a thesis in Quantum Optics.
In other words, the thesis should be aimed, in my opinion, at active researchers who want a quick and painless introduction to the author's field. The other category a thesis should be aimed at would be researchers who want to learn the methods and techniques used to obtained the results presented in it.