QGX: Performance issue when many starting branches are found

[tokebe]

The below query graph from ARAX (and similar queries with nodes that expand to many subclass descendants) seems to cause Retriever to chug. This is odd, because hundreds of nodes being found for an intermediate (or really any non-starting) node seems not to cause problems.

{
  "nodes": {
    "SN": {
      "ids": null,
      "categories": [
        "biolink:ChemicalEntity"
      ],
      "set_interpretation": "BATCH",
      "constraints": [],
      "member_ids": [],
      "is_set": true
    },
    "ON": {
      "ids": [
        "NCBIGene:675"
      ],
      "categories": [
        "biolink:Gene",
        "biolink:Protein"
      ],
      "set_interpretation": "BATCH",
      "constraints": [],
      "member_ids": [],
      "is_set": false
    }
  },
  "edges": {
    "t_edge": {
      "subject": "SN",
      "object": "ON",
      "knowledge_type": "lookup",
      "predicates": [
        "biolink:affects"
      ],
      "attribute_constraints": [],
      "qualifier_constraints": [
        {
          "qualifier_set": [
            {
              "qualifier_type_id": "biolink:object_aspect_qualifier",
              "qualifier_value": "activity_or_abundance"
            },
            {
              "qualifier_type_id": "biolink:object_direction_qualifier",
              "qualifier_value": "decreased"
            }
          ]
        }
      ]
    }
  }
}

[tokebe]

Turns out this isn't much of a performance issue, but changes could still be made (particularly to the batcher) to reduce log bloat and presumably speed things up a bit.