Open Stack

The current design for ovs-neutron-agent is that it will wipe out all 
flows configured on the system when it starts up, recreating them for 
each neutron port it's aware of.  This has a not-so-desirable side 
effects that there's a temporary hiccup in network connectivity for the 
VMs on the host.

My questions to the list: Is there a reason it was designed this way 
(other than "Everything on the system must be managed by OpenStack")? 
Is there ongoing work to address this or would it be a worthwhile 
contribution from our side?