Open Stack

I was looking over the admin guide http://docs.openstack.org/admin-guide-cloud/networking_config-agents.html#configure-l3-agent and noticed this:

"If you reboot a node that runs the L3 agent, you must run the neutron-ovs-cleanup command before the neutron-l3-agent service starts".

Taking a look at neutron-ovs-cleanup, it appears to remove stray veth pairs and tap port in OVS.  The admin guide suggests ensuring neutron-ovs-cleanup runs before L3 agent and DHCP agent start when rebooting a node.  My question is whether there is something special about a reboot vs. an agent restart that is the genesis of this note in the admin guide.  What conditions can get you into a state where neutron-ovs-cleanup is required?  Is it just a matter of the OVS agent getting out of sync and needing to go back to a clean slate? Can anyone shed some light on this note?

-Ryan