The chunks helper for concurrency workloads has a function to retrieve the primary replica set node from a shard.

If this function doesn't find a primary node from any of the current nodes, it returns immediately. This is not stepdown-proof, as a node could be transitioning to primary but not yet reporting isMaster: true.

We should use assert.soon with a short-ish timeout instead of immediately returning and failing a test, similar to what the replica set test class does.