Create JS tests and infrastructure to validate that:

• A recording can be started on a single mongod node
• File(s) are created
• Mongor (or running equivalent function from a native method) is capable of replaying the files, issuing commands against a single mongod node

It should expose information on:

• When a command failed during replay (to confirm success, or to allow negative tests)

It could include validation of the resulting dataset. This could be as simple as running a colscan on both the original and "shadow" cluster, and providing "fuzzy" comparisons which can ignore unimportant changes (e.g., auto-generated fields like object IDs).