Simulate load

Push several tickets through the pipeline at once and watch them advance independently.

Groq's free tier caps the specialist model at 8k tokens/minute, so runs are throttled to a small concurrency and retry automatically on a rate-limit instead of failing. On Groq's Dev tier the whole batch runs fully in parallel.