We should add a new perf test that runs dbCheck in the background. The data in the replica set should consist of documents that generate same-value index keys, to test the performance in the situation where secondary index checks have to run a large batch due to same-value index keys.