The existing shard split performance test is written as a simple jstest, which runs as part of the serverless jstest suite. Since the test is not run in an isolated environment, we periodically receive BFs that the critical section took longer than 500ms to complete. This ticket will cover converting this test to something we can run in DSI with guaranteed isolation from other tests.

We have an existing test for moveChunk which we might model our work after, though it is an open question about how to enforce a hard upper bound on 500ms.