I am intending on designing a MongoDB Atlas cluster with a typical three node replicaset that are all electable as primary. Then adding in a couple of readonly analytical nodes for offloading heavier query workloads.

I am wondering if I can configure an aggregation pipeline in a way that selectively queries the readonly analytical nodes for input then still outputs or merges the results into the regular primary electable replicaset nodes in order to isolate the query load to the readonly nodes.