Design a lab machine fleet monitor with remote reboot

Design a service that watches a fleet of lab machines running compute workloads. Each machine polls the service for work; a work item names a payload the machine must download, verify, start and then keep an eye on. A central dashboard shows the state of every payload and machine.

Cover the work-assignment protocol, how the service notices a machine or payload that has hung or gone silent, how an operator (or the system itself) reboots a stuck machine remotely, what state lives where, and how the design survives machines and the service itself failing.