There are totally 8 nodes in the config. By looking at the log, it seems that when we do reconfig, node 7(20547) is way behind others in the reconfig process. In the test, we move forward based on whether the reconfig succeeds on majority of nodes, so one slow node doesn't affect the test process, however, when we shutdown the replica set, we do rely on every node to finish the reconfig process or the shutdown will error out, leading to test failure.

We can add a configuration check for every node to be up-to-date before shutdown replica set, so the test won't fail on replica set shutdown.