Before running the first aggregation query, and immediately after waiting for replication, clear the operation metrics.

If a secondary is slow to replicate the drop of "newDB" in the test, then it will be reported in the metrics earlier on than it should, causing the test to fail.