While we're currently storing results as if they were independent of other results, this is not always the case and is a known limitation of the current resultsdb.

At Flock 2013, Josef and I talked about possibly migrating ResultsDB's backing datastore from a RDB to something that supports more unstructured relationships (like a graph database).

This should allow for better linking between results, other results, packages, builds and testcases without needing an overly complex schema and lots of join queries. However, this would not be a trivial migration and needs to be investigated before we seriously consider any migration. Eventually, a small proof-of-concept would be the product of this investigation if it appears to be an idea worth considering.