Burst compute is a second rental and a copy

The eval grid needs a weekend of compute your box doesn't have. On burst platforms, that means rebuilding your environment inside their pods and watching a meter.

On Hoody, it means the marketplace. Rent a larger box for the duration: pricing is flat per server, by spec and term. Then POST /containers/{id}/copy with the new server as the target: the worker container lands there whole, with its packages, data, and configuration. Fan the runs out, copy the results container back home, and let the rental end.

Your environment never got translated into someone else's job format. It moved, ran, and came back as the same machine.