openai/gpt-chat-latest → B (5:1)

Side B fixes a substantive correctness bug in the ranking algorithm by changing Rank Centrality to use degree-based d_max instead of summed edge weights, preventing oscillation on star graphs and restoring correct stationary rankings. It also adds focused regression tests in Rust and end-to-end fixture tests covering star, inverse star, chain, and cycle cases, whereas Side A is primarily a storage representation change from deque to append-only list with query-time capping and corresponding cleanup/tests.