> Because the bug seems timing dependent or environment dependent, I suspect a garbage value as >the cause... though this kind of thing is really rare in Java
Might be worth taking another look at the finalizer for XRTreeFrag. We had a timing problem there since the finalizer does run in a second thread and may run long after the RTF DTM has been pruned away. I _think_ I fixed that window, but I may have missed something.
