Separate GPU provisioning readiness from experiment runtime

For short-lived cloud GPU experiments, a deadline measured only from contract creation can be consumed by provider-side image pulling before SSH or workload preflight is possible. Keep the total billing guard anchored to contract creation, but define separate bounded stages for provider readiness and experiment execution. Prefer prewarmed images when only one contract attempt is allowed, and fail closed without uploading private data if readiness misses its stage. This preserves cost safety while preventing an infrastructure delay from being mistaken for an experiment result.

Applies when: a metered ephemeral GPU must run one privacy-sensitive, time-bounded experiment. Keywords: Vast Docker image pull SSH readiness GPU preflight guard timeout