Customers have suffered major downtimes during live-migration with bridge CNI. This was caused by:
1. Auto-configuration of IPv6 address on the veth side in container netns is done during target pod setup
2. This issues neighbor solicitation advertising a new source of the MAC assigned to the veth (before it is reassigned to the guest)
3. This confuses network hardware into thinking that the MAC is now available on a different part of the network - breaking connections to the source, even before the migration is completed

I suspect we have the same issue happening with localnet.

4.16

Needs to be checked

1. See the reproducer of https://issues.redhat.com/browse/CNV-28040
2. Run LM and listen to neighbor solicitation
3.

I suspect there will be neighbor solicitation packets

There must be total silence from the destination until the VM completes LM