My point is that much more intelligent, useful and efficient queries could be possible by making use of the fact that code is a data structure, and not just text. Various things could be inferred automatically, which would be especially handy in legacy C source code where it's really hard to make head or tail of (I see a lot of that in my job).
It should be able to answer questions like "where does this value come from", "how is it computed" and show a tree/chain of statements. Or "where is this structure accessed" ... and so on.
As clang makes these things easier, I hope there will be a new resurgence in intelligent code comprehension tools. Those don't necessarily have to be point and click, but could have different UIs...