How do Elasticsearch metric/matrix aggregations execute internally? What happens if the query includes filter (SQL "where")? Does the execution plan follow the alg
