Naomi,

I do not understand the algorithm at all, I do not know what this code is trying to accomplish, but what about working more with the data, wouldn't it be faster?, for example at the beginning you have

* Find a root node in the TreeView
	oTree = .TList2
	if oTree.Nodes.count > 0
		for i = 1 to oTree.Nodes.count
			oTmp = oTree.Nodes(i).parent
			if isnull(oTmp) then
				iTmpIndex = i
				exit
			endif
		next i

Which finds only one of the possible multiple root nodes of the treeview, wouldn't be faster to do

select node_id from ItemTree where parent_id is null into cursor c_roots...

You might even be able to avoid the next step, finding the sibling if you use order by sort_order, I think. Anyways, what I am trying to say is that maybe it is better to work with a mix of data instead of just scanning the tree which might be expensive.