# Changelog

## Score

- CAI 55 → 55 (-0.6)
- Rubric changed (rubric-2026.10.5 → rubric-2026.11.2) — scores are not directly comparable.

## Lenses

- Code Health 61 → 61 (+0.0)
- Architecture 96 → 96 (+0.0)
- Maturity 42 → 39 (-3.6)
- Readiness 64 → 64 (+0.0)
- Security 77 → 91 (+13.4)

## Resolved (3)

- Duplicate intent between Event methods and Debounce helper methods. Event.upsert?() and Event.destroy?() seem to determine the type of operation for an event. Debounce.upsert_event?() and Debounce.destroy_event?() perform the same logical check but are located in a helper class. This duplication suggests that the Event class should be the single source of truth for determining the operation type, and the Debounce class should delegate to it or the Event class should expose these predicates directly.
- Inconsistent naming and scope for data retrieval operations. ORMAdapter.Base.find() suggests a generic lookup operation, while Receiving.Model.*.find_and_save() implies a specific transactional write-after-read pattern. If find() in the adapter is meant to be the generic lookup, it should not be mixed with the specific 'find_and_save' logic in the receiving models without clear distinction. Furthermore, 'find' is a very common name; if it behaves differently in different contexts (simple lookup vs. part of a save transaction), it should be named more descriptively (e.g., lookup_by_key vs. find_and_lock_for_update).
- Redundant entry points for publishing/syncing. TableSync.sync() appears to be a high-level convenience method that likely delegates to the Publishing classes. However, the Publishing.Single and Publishing.Batch classes expose their own publish_now/publish_later methods. This creates confusion about whether a user should call TableSync.sync() or instantiate a Publishing object and call publish_now. The intent (triggering a sync/publish) is identical, but the API surface offers two distinct paths with different signatures and object lifecycles.
