Comment 30 for bug 1596075

I think I've tracked down the issue to agent-related queries causing DBDeadlocks that happen after the subnet creation (in this particular case updating the agent load value). This causes the API layer to use its retry mechanism even though the subnet has been persisted to the DB, which leads to the overlapping error.

Please try again with the latest version of https://review.openstack.org/#/c/339226/. It has a crude fix, but it should cause the issue to go away to confirm my suspicions.