In the new createCollectionCoordinator, hitting a ShardNotFound error triggers some updates to the list of involved shardIds. This prevents us from continuously trying to contact or otherwise involve a shard the has been removed.

However, the _cleanupOnAbort function which is triggered by calling triggerCleanup does not have any onError handlers. This means that if an involved shard is removed while the create coordinator is running cleanup, the coordinator would continue to try to release the critical section on this removed shard.