@adelgado What an interesting idea. Search "instances," each an aspect of your searched index, with groups of sources aggregating, annotating, and maintaining lists of resources of interest/importance to them.
There would be lots of overlap but that would be good because you use that information for weighting results.