It's weird that this bug report has been rejected, but I guess that's why it has remained this way for many years. I would guess there are two factors that have gotten Eclipse into such a strange stable state around this problem.
The first factor is interface mania. File search is an obvious feature if you discuss it by itself, but the implementation goes through Eclipse's resources API. Thus, every discussion about file search brings up design questions about the resources API, which for sure nobody wants to touch at this point. This is a cost of reuse and abstraction that make it not as good of a deal as it seems like on the surface. In the case of Eclipse's file search, maybe they should have it avoid the resources API for the main search functionality.
The other factor is Stockholme syndrome. You see it all the time if you work in one development group and then in another. You'll see something work fine in the first group's software, only to be borked up in some horrid way in the second one. It can be hard to fix such problems. The people in the second group are often so used to the problem that they start attributing it to something fundamental and special about their system, much like a kidnapping victim. If you start discussing how it could be fixed, they think you are talking science fiction, switch to blue-sky mode, and start throwing out all kind of whacky ideas.