This is related to a build failure on non-document level locking storage engines.

The call to OpObserver::preRenameCollection here will get the locks on the Oplog and maintain them during the duration of the WriteUnitOfWork. The check for assertNoIndexBuildInProgForCollection() requires the IndexBuildsCoordinator _mutex. But if we have an operation that is already holding the _mutex and is waiting for the Oplog locks, we will deadlock.

To solve this, we should check for background operations prior to calling preRenameCollection().