The FCV 4.4 migration protocol involves doing a retryable write against the recipient shard, and retryable writes are not supported on storage engines that do not have document-level locking.

This ticket is to check whether we still support running shard servers on storage engines without document-level locking.

If it's not supported, the ticket should figure out what to do about sharding tests that do migrations with the ephemeralForTest storage engine.