What do developers want from a source code search engine?

Like many developers out there, I find it unfortunate to read about the upcoming shutdown of Google Code Search. In reaction to this, several developers (including those here at HN) indicated the value of having a open source code search engine such as Google Code Search and Koders.com. There have also been a couple of comments about the usefulness and challenges in using these tools. On the other hand, after doing years of research in understanding code search work for developers (http://goo.gl/GKURw), and having recently started working in the Koders's engineering team, I am very interested in the recent reactions.

It is great to see when developers get excited about these topics in places like HN. In our code search development team, we are already focusing on solving problems that developers pointed out recently in the discussions (for example, non-obvious ways to searching for identifiers with symbols, search performance etc). Some of these problems will be solved in the very near future. The site was already scheduled to go through a major upgrade to the latest version of our technology. Beyond that we are taking a step further by thinking deeper about how a code search engine should be different. For example, one of our current focus areas is leveraging more structural and relational information in code to improve code search capabilities. The innovation possibilities, and thus the challenges in getting them right, are abundant. However, it would be awesome if developers out there would speak to developers like us in the Koders team about what they'd like to see in a code search engine so that we can provide the best code search experience for developers.

So, here is a question for developers out there wishing for a better code search engine: What would a killer code search engine do for you ?