The PeriodicShardedIndexConsistencyChecker runs an aggregation on secondary nodes of a shard. This secondary read can trigger a refresh if the secondary doesn't know the filtering information for the database.

The read_committed_lookup test disables replication on both secondaries of one of the shards (one via a fail point and one because it is an arbiter) which will make the primary unable to refresh the catalog cache due to SERVER-78115 which introduced a majority write to ensure the primary is in fact primary.

The test runs a lookup on the primary which will require the routing information for the collection we are doing a lookup from, which will get stuck behind the refresh which is waiting for majority write concern.

Because this will be fixed by splitting the routing and filtering caches into two during SCAAR (SERVER-84243) and fixing this in the Shard Server Catalog Cache Loader would mean introducing a lot of logic for database refreshes (which we expect to be not very frequent), we will just fix the test to prevent the background refresh and wait for 8.1 to fix this.