Thanks Alex.  For your interest, and anyone else in dealing with recursive
boms, nodes etc:

I only load my node structures one level at a time.  Each node that gets
added, also tests to see if there may be nodes below it, but only at the next
level below and no further.  If there should be child nodes at the next
level, then I simply add a single dummy child node so that the + sign
appears.  You can do this by returning a data set of the required table using
an outer join to the table below it.  Group the data and use
CountOfTablebelowfield to test if there should be child nodes.

       If MDLSset!CountOfChild > 0 Then
           Set nodCurrent = objTree.Nodes.Add(nodCurrent, tvwChild, ,
"Dummy")
       End If

Then, on the event for node expanding, I delete the single dummy node,
populate with the "real" next nodes and check once again for child nodes
below them adding a dummy again if required.  And so it goes on....

   objTree.Nodes.Remove currNode.Child.Index
   Set MDLSset= CurrentDb.OpenRecordset("SELECT SSTH.STID, SSTH.STDSC,
Count(SSTD.SDSQ) AS CountOfSDSQ.....
   Set nodCurrent = objTree.Nodes.Add(currNode, tvwChild, , strText)
   If MDLSset!CountOfSDSQ > 0 Then
       Set nodCurrent = objTree.Nodes.Add(nodCurrent, tvwChild, , "Dummy")
   End If

The only downside is that when you collapse your node, you must delete all
nodes and add a dummy node again (unless you check on expand not to add nodes
if the first node is not a dummy node).  Which means if you have expanded
lower levels, and collapse a higher level, you have lost all your lower level
expansions if you re-explode again (which I personally never want anyway).

Doing it this way means you only load what you want, and you always have the
latest values and do not need to refresh  - each node exploded gives you the
latest data.