### Note on root cause

The behavior below is verified three ways: by reading the merge code, by executing the real proxy merge path (`add_litellm_data_to_request`) in-process against constructed key/team/org auth objects, and by reproducing it end to end against a running proxy, including a full precedence matrix read back from `/spend/logs/v2`. The organization-level drop is not inferred

### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate

Closest items found, all distinct from this one: - #29054 and #26149 concern preserving the request-level `x-litellm-spend-logs-metadata` header on `/v1/messages` and pass-through routes; they are about the request path, not organization-level collection - #31784 (merged) exposes `project_alias` in Prometheus custom labels; it touches the org/project/team/key hierarchy but not `spend_logs_metadata` merging - #31867 / #31875 concern tags and `session_id` forwarding in cascaded-proxy topology, a different field and code path

### What happened?

`spend_logs_metadata` is collected and merged from the request, the API key, and the team, but not from the organization, even though `or…