When startup completes, it's not guaranteed that the node is ready for another reconfig because we update rsConfigState to kConfigReady after[ startup completes|https://github.com/10gen/mongo/blob/f029383f47c9a2217caa47dc2f8dbde93babd977/src/mongo/db/repl/replication_coordinator_impl.cpp#L579]. 

That says, there is a race condition in the test that the restarted node can reject the reconfig if the node hasn't become kConfigReady. We should wait the restarted node to become secondary before we issue the reconfig.