The DropDatabaseCoordinator inserts a document on kConfigDropPendingDBsNamespace after SERVER-96828 to notify the delete to a concurrent CreateDatabase operation.

The document inserted is deleted at the end of the coordinator here.

In a multiversion scenario with failovers, we may end up not removing the inserted document. For example:

1. A DropDatabaseCoordinator starts on a node with v8.1. A document is inserted on kConfigDropPendingDBsNamespace.
2. There is a step-down/step-up in the middle of the dropDatabase operation, and a node with v8.0 becomes the primary.
3. The new primary resumes the DropDatabaseCoordinator operation and doesn't delete the document inserted on kConfigDropPendingDBsNamespace.

Keeping a document on kConfigDropPendingDBsNamespace will cause that:

• The next time we run a DropDatabaseCoordinator on the same database, we'll hit a DuplicateKey error and the coordinator will be retried forever.
• After running setFCV to the latest version, a CreateDatabaseCoordinator execution will fail on this check and will retry forever.