We have many query tests which need to do something special with a ShardingTest - often just sharding a collection and ensuring it's on two different shards or something like that.

I think we should invest in a way to have these tests share a single sharded cluster, to dramatically speed up the tests and also de-clutter the logs. Each test could have a unique collection name or db name, similar to jscore. Unlike the sharded jscore passthrough, it would be expected/encouraged for tests to modify the sharding distribution of their datasets.