There's a lot of interest in an open search engine; I'd like to throw an idea out and see what people think.
Given that people have areas of interest that are more important to them (and a great deal of areas that they don't care about) it may be possible to enhance the search experience (& privacy) by downloading data on selected verticals to the users machine (ideally to an SSD). Queries in those areas could then be run mostly/entireley locally (possibly with more exhaustive processing too). This would also allow for a great deal of testing to be done by technical users on optimum settings.
If you needed to set aside 250GB of space on an SSD in exchange for better/alternative search results (in areas you really care about) would you make that sacrifice? Would you be prepared to go further and dedicate an entire, expensive, SSD to the task (such as 1TB). Clearly, given bandwidth issues, the refresh rate of data would be lower than mainstream search engines so this would work well for info that changes less frequently (although refreshing data would get better over time).