Treat post-validation SCM errors as terminal: add RunStatusValue import and SCM error prefixes, introduce _escalate_after_scm_failure helper to mark a story ESCALATED and update run/story status to halted, and change commit_node so commit_story/remove_worktree failures escalate instead of being best-effort. Move durable run/story updates to after critical SCM ops. Update halt-reason and suggested-fix logic to surface scm_error and provide user guidance. Add tests covering commit and cleanup SCM failures and finalize_node behavior.