CareerCup

sounds about right in your english explanation, except you aren't adding a new null to the end of the queue, so this code will never be able to determine the end of the level after the root level :P
The loop is adding the left and right to the queue only if it is not null, which will reiterate in a recursive fashion to repeat adding non-null lefts and rights, which means you are never adding an extra null.

Iterative method using Queue:
Aux storage:
Queue traverseQueue
List<LinkedList> or LinkedList[maxLevelOfTree]
int levelIndex

Psuedo code:

traverseQueue.add(root);
levelIndex = 0;
while (traverseQueue.Size() > 0) {
	LinkedList head = new LinkedList();
	LinkedList current = head;
	for (int temp = traverseQueue.size(); temp > 0; temp--) {
		TreeNode temp = traverseQueue.remove();
		current.value = temp.value;
		for (TreeNode child : temp.childs) {
			traverseQueue.add(child);
		}
		if (temp > 1) {
			// if not the last node of current level
			current.next = new LinkedList();
			current = current.next;
		}
	}
	listOfLinkedList[levelIndex] = head;
	levelIndex++;
}
return listOfLinkedList;

I think this is a question about data structure and OOP design..

Each tweet would be a tree node, with an original tweet being the root.

Each node contains a tweet object, which holds list value of favorites, a formatted string, a bitmap image, and a list of people in the conversation ("@person 1", "@person 2") and the child nodes may be retweets (which are also tree nodes).

Then we may have a linked list of tweets, which is the replies (conversation)

All tweets, retweets, and replies are tweet objects which is all in a giant map that has the tweet ID as a key and tweet object as value..

I don't use twitter so these are just ideas out of my brain :)

I'm sure there are different structures, such as an abstract tweet object with child classes such as retweet, reply, and root tweet, which holds different properties.

I guess it all comes down to what you personally think is the best structure.