ttl_operation_metrics_multi_dbs.js may fail due to unexpected metrics collected on the second DB if the TTL monitor is not able to run two passes over the TTL index in db2.coll. This may happen on a slow machine and the TTL index is not registered in time for the first pass to run.
The test currently uses a fail point that pauses TTL operation in the middle of a TTL pass. We could consider using a different existing fail point that allows us to pause between passes so that we can get a complete view of the catalog when the test resumes TTL operation.