I can play with that idea.  Is there an example?  It seems like the ultratree would have to auto-magically know to switch from one relationship to the other at the second level of the hierarchy. 

Otherwise is there any way to pre-emptively manage the visibility of the nodes before initialization (ie. outside of the "InitializeDataNode " event)?  ... or is it possible to manage that in an async way that is happening outside of the u/i thread?

Can you estimate for me how many nodes is reasonable for an ultratree?  I assumed I had already hit the limit, but even if get over the current hump (with UltraTreeNode::set_Visible), I can't imagine that the ultratree will allow me to add nodes indefinitely.

Also, is it possible to upload attachments to this forum, in case I modify my example and it still isn't working.  Are attachments disabled for me?

UltraTree binds to the DataSource you give it. So it doesn't have to switch anything. The DataSource would alraedy have the relationships defined. It's just a question of building your DataSet properly. 

It would be hard to provide a meaningful example of this, since you are presumably getting your data from a SQl Server or something that we don't have access to. I could build a DataSet with the appropriate relationships in code, but that probably wouldn't help you very much since your real application is different and gets data from a server. 

The number of reasonable nodes is arbitrary. There's no hard and fast rule. It depends on the speed of the hardward and the amount of memory and such. You might want to think about how many nodes is reasonable for your UI, anyway. No human user could possibly deal with 10,000 nodes in a tree in any meaningful way. Maybe you need some sort of search or query interface so the user can pull down the nodes they need to work with instead of the entire set. 

Anyway, the approach is not that complicated though.

Right now, you have a DataSet with a single table and a relationship to itself. Presumably every row in your data has an Id that uniquely identifies the row and also a ParentId which indicates the parent row for that row. And some of your rows have Null or 0 as the ParentId, indicating that they are root-level rows with no parent. 

Instead of doing that, you need to create a DataSet with 2 tables. 

You get the root table using a query that only returns root-level rows. So you use a Select or something and put in a condition that the ParentId is null. Let's call this table "Root Table"

Then you get a second table which has the opposite condition and only get the rows where the ParentId is NOT null. Let's call this table "Child Table"

Once you have those two tables in your DataSet, you create two relationships. 

        private DataSet GetRecursiveDataSetWithTwoTables()
        {
            DataSet ds = new DataSet();
            var parentTable = CreateRootTableWithData();
            parentTable.TableName = "Root table";
            ds.Tables.Add(parentTable);

            var childTable = CreateChildTableWithData();
            childTable.TableName = "Child table";
            ds.Tables.Add(childTable);

            // We have to specify false here for CreateConstraints, because some of the children have 
            // parent that are in a different table. 
            ds.Relations.Add("Parent To Child", parentTable.Columns["Id"], childTable.Columns["ParentID"], false);            
            ds.Relations.Add("Child To Child", childTable.Columns["Id"], childTable.Columns["ParentID"], false);
            return ds;
        }