it's hard to analyze exactly what's going on without access to the sample data, but I suspect there's some aspect that's making the rules fire multiple times (they execute as non-lazy subqueries) and therefore it would be repeatedly running the same lucene searches

you could try adding some debug clauses to the and legs to help figure things out, like: [(println "running client search") ?ignore][(nil? ?ignore)]

> is it possible to achieve the same speed with single query? you could fold those top-level queries into a single query as explicit subqueries. There's not a huge advantage to doing so (it doesn't get much easier to read), but it means you only have one single query.