When a VM is accesed directly from an external network the TCP traffic does not get SNATed until TCP connection starts pushing data. This causes TCP reset.

Customer setup /Bug report:

https://bugzilla.redhat.com/show_bug.cgi?id=2263502#c0

Reproduced here(shows tcpdump):

  https://bugzilla.redhat.com/show_bug.cgi?id=2263502#c3

It seems there is a patch for this exact scenarion upstream

https://patchwork.ozlabs.org/project/ovn/patch/20240207155627.806188-1-martin.kalcok@canonical.com/