SERVER-43758 added a test to check that a node would successfully apply pending oplog entries with -shardsvr and -queryableBackupMode flags. In order to do so, it manually stops a secondary node of a shard, and then starts a custom mongod with said flags to verify the oplog entries were applied on startup. The problem with this approach is that after shutting down the test node any correctness validation performed in the stop of the ShardingTest object will fail.

We should do something similar to the read only tests set up, simply restore the node that was shutdown from the cluster, and allow the correctness checks to work as expected.