The idea would be to trigger an un/clean shutdown on the donor shard primary or the recipient shard primary during the resharding operation and verify it doesn't affect the recipients' data eventually becoming strictly consistent with the donors' data.