openai/gpt-chat-latest → B (2:1)

Side B changes the sibling navigation logic so each unranked sibling becomes its own navigation group instead of lumping all unranked items together, matching the documented behavior and adding a regression test that verifies the grouping. Side A mainly adds CLI presentation of already-existing connectivity statistics with a formatter and tests, improving visibility but not changing the underlying project behavior.