Comment by ge0ffrey

Ideally, the solver doesn't just run on the backend, it's also orchestrated to run a pod per solver, with scale to zero.