We tested out Groq with Llama 30B and saw an increased in inference speed, but the API was unreliable, sometimes not returning anything, which for us is a non-starter.<p>I am curious if anyone is using Groq in a production environment and if so, what their use case is?