HSBC ran into an issue where they wanted to move a Replica Set from an old Ops Manager environment to a new one.

They stopped the Replica Set, completely unmanaged it from Ops Manager (which deleted the binaries in the bin directory under the Agent) and then tried to import the Replica Set as an existing instance to their new Ops Manager environment.

As expected, the import failed because the Replica Set was not running.  Nor could they manually restart the Replica Set because the binaries were no longer available on the host as they had not manually installed MongoDB.

Although it would seem self-evident that a Replica Set must be running and healthy in order to import it into Ops Manager, there is no explicit note/warning in the documentation or the UI to that effect.

Even though it is an edge case, it might be worth adding a note to the docs that stopping a managed deployment and completely removing it from Ops Manager will delete the binaries under the MongoDB Agent bin directory.  There should also be a note that when importing an existing Replica Set to Ops Manager that the instance must be running and healthy for the import to be successful.