Codex performance may degrade under reasoning-token clustering in GPT-5.5

A developer flagged potential performance issues in OpenAI's Codex when using reasoning-token clustering in GPT-5.5. The finding suggests that certain architectural changes could be creating bottlenecks for code generation tasks—worth investigating if you're building on top of these models.