Parallel secondary oplog application creates insertion-order noise that can make the monotonicity computation done by analyzeShardKey incorrect. This means that the tests evaluating the monotonicity computation can occasionally fail due to this noise, an issue we've seen recurrently. While we have tried to fix the issue by running with larger (more production-like) document counts to reduce noise, these makes the tests run longer and has not completely fixed the issue. We could consider other approaches like slowing down the insertion rate to avoid parallel application, but this again would make tests run very slowly for questionable value. 

At this point, we think we should just stop doing the monotonicty check on secondaries, which has limited value and is causing this noise. An alternative to consider would be setting the replWriterThreadCount to 1 on secondaries, so we can still test on secondaries but avoid the noise induced by parallel application.