Design a power plant telemetry and analysis platform

Plant operators watch telemetry from generating units and run analysis jobs over a chosen time range to investigate anomalies. Design the platform: ingesting and serving recent telemetry, submitting jobs with an idempotency key, running them across workers with leases and checkpoints so a lost worker's job resumes elsewhere, and publishing exactly one final result per job even when work is delivered twice. Operators should see each job's status, inputs, failures and results. Physical control of the plant is out of scope.