How does this compare to Fly.io which AFAIK wakes on requests in a very similar way with containers made into Firecracker VM's (unless using GPU, I think)? I su

How does this compare to Fly.io which AFAIK wakes on requests in a very similar way with containers made into Firecracker VM's (unless using GPU, I think)? I suppose Fly typically doesn't need to scale per request so has a typical cheaper max cost? I guess you'd just set the concurrency but how well that scales, I don't know.