PXD-580 ⁃ Flakiness in GraphQL responses

[david4096]

Try repeatedly making this request to the GraphQL interface:

    {
     submitted_aligned_reads(project_id: "topmed-public") {
        id
           data_format
      data_category
        file_name
        file_size
        md5sum
        submitter_id
        updated_datetime
        created_datetime
       _links {
          id
          type
        }
      }
    }

The second or third time you might observe an error where the links are not populated and a graphene message propagates to the client.

      {
        "_links": [
          null,
          null
        ],
        "created_datetime": "2018-05-26T13:37:15.281868+00:00",
        "data_category": "Sequencing Reads",
        "data_format": "CRAM",
        "file_name": "NWD905240.recab.cram",
        "file_size": 20699570093,
        "id": "ff1739f7-8da7-41e1-9dd7-2bb1c1278059",
        "md5sum": "7af0014daca79bcc4bbed32fc128c066",
        "submitter_id": "NWD905240-cram",
        "updated_datetime": "2018-05-26T13:43:10.978473+00:00"
      }
    ]
  },
  "errors": [
    "The type read_group does not match with the associated graphene type read_group.",
    "The type aligned_reads_index does not match with the associated graphene type aligned_reads_index.",

uc-cdis/sheepdog#142

[zflamig]

This feels like a race, but we are using the syncexecutor graphql-python/graphene#314

Need to think/poke at code some more