When upgrading to OSP 13 from previous versions, supporting processes for neutron agents like dnsmasq, haproxy, keepalived, etc. created prior to upgrade continue to run on baremetal until the node is rebooted. New processes instantiated for new neutron objects (e.g. routers, networks, etc) will run in containers resulting in a mix of both. This has potential implications with respect to support and troubleshooting.
To clarify: this should not impact the upgrade itself from a user point of view. As indicated in comment #0, the real impact is on the operator only.
This was an initial potential issue when switching to containerized neutron, but since then all upgrades went smoothly and neutron can shut down baremetal process and then start sidecar container ones (we base process control on PID). So we can close this one now (almost 3 releases after switching to containerized Neutron)