Given the CI failures that were observed during initial testing of running RHCOS 9 in OCP (see https://issues.redhat.com/browse/OCPBUGS-7275?focusedCommentId=21705994&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-21705994), the MCO will need to detect this upgrade and do some preliminary steps so that the upgrade can succeed. This is due in part to a newer version of rpm-ostree needing to be present due to https://github.com/coreos/rpm-ostree/pull/4302.

From the MCO side, it seems like we'll need the following:

• Detect that an OS upgrade is happening.
• Detect that the OS version is changing from 8.6 -> 9.0+.
• Pull the RHCOS 9.0+ extensions container, extract a newer version of rpm-ostree from that container, install it on the node, then restart the rpm-ostree daemon (e.g., systemctl restart rpm-ostreed).
• Once the rpm-ostreed daemon restarts, the OS upgrade can proceed as normal.