It appears that this test, when run in the sharding with balancer suite on a slower debug variant can get stuck behind sharding operations. There is a server parameter that can increase the retry count to reduce the chances of this occurring until the Shard Role API is ready.