openai/gpt-chat-latest → A (3:1)

Side A repairs multiple concrete failures in the OAuth test infrastructure: it fixes query parsing (`str/split` regex), reads POST bodies correctly (`getRequestBody`), avoids null token/state crashes, adjusts redirect handling, wraps handlers to prevent server crashes, and improves Playwright test synchronization and selectors. Side B is a focused behavioral improvement that skips pinned Reddit posts during import with a helper and regression test, but its scope and impact are narrower than restoring and hardening the end-to-end authentication test framework.