The parallel algorithms have issues if seeds have duplicate IDs.
(Need to add code at init time to check for this).