openai/gpt-chat-latest → A (4:1)

Side A fixes a core correctness bug in the ranking algorithm by switching Rank Centrality to the canonical degree-based d_max, eliminating oscillation in star topologies and adding focused regression tests (Rust and end-to-end fixture tests) that verify the corrected behavior. Side B is a broad URL/routing refactor that centralizes room path handling and adds URL normalization utilities, but it is largely structural and API reshaping rather than fixing a comparably fundamental algorithmic correctness issue.