Right now there's an approx 15s delay when playing media due to k8s scheduling, launching the pod, mounting volumes etc.

We could investigate improving this time by keeping some 'warm' transcoders. This would require a bit of an architectural change, but is absolutely possible.