Comment 4 for bug 1786346

I've got a test patch here:

https://review.openstack.org/#/c/553608/

It drops the vif_plugging_timeout to 30 seconds in a multi-node linuxbridge job (so we run live migration) and sets live_migration_wait_for_vif_plug=True so the compute manager will wait for and consume the network-vif-plugged event before the driver.live_migration method runs and adds its own waiter. Hopefully with the 1 second bw toggle on the guest that will exceed the 30 second vif_plugging_timeout threshold and we'll see the live migration fail.