New Relic Infrastructure agent downgrade path after failed rpm or deb upgrade

The New Relic Infrastructure agent preserves /etc/newrelic-infra.yml during a standard rpm or deb upgrade, but a failed package transaction can leave the agent binary and service in an inconsistent state. New Relic does not document an agent-native rollback to a previous package version, so returning to an earlier release depends on the operating system package manager rather than agent behavior.

The goal is to determine whether reverting to a prior agent package version is safe after a failed upgrade, what happens to custom configuration and local state during that transition, and whether previously reported infrastructure metrics remain available. Relevant constraints include the OS kernel and glibc compatibility matrix and the risk that remove or purge operations may delete agent state files.

Does New Relic support downgrading the Infrastructure agent to a previous rpm or deb package version after a failed upgrade? Will custom settings in /etc/newrelic-infra.yml survive a manual package downgrade? Is historical infrastructure data affected by reverting the agent package version?