A query against system:completed_requests needs to fetch the completed_requests entries from all query nodes in the cluster, and then perform further processing on all these entries on the query node executing the query.
If there are a large number of completed_requests entries across nodes, the query can take a longer amount of time.
If you are in a multi-node set up and only require entries from particular nodes, you can add a predicate on the “node” field. This can make the query faster.
Example: select * from system:completed_requests where node = "node_name";