Original practice scenario: A hypothetical semiconductor-equipment service sends a diagnostic command to a machine. The acknowledgment is lost, and repeating the command might alter state. Define the recovery protocol.
Read model answer +
Classify the command by its effects before deciding whether it can be retried. A missing acknowledgment creates uncertainty about execution, which should be resolved through durable command status where possible.
Work through it
Give the command an identity, payload fingerprint and authorized target generation; the machine records acceptance and completion evidence under that identity.
Distinguish read-only diagnostics from state-changing operations, defining duplicate handling and preconditions for each command class.
After communication recovery, query status before resubmission, reconcile device and service records, and escalate an unknown effect instead of assuming failure.
Tradeoffs to discuss
Durable local receipts add storage and protocol complexity but reduce ambiguous retries; pure status polling may delay service recovery.
Common pitfalls
A device restart can invalidate an old target generation or erase volatile evidence.
This practice protocol must retain separate operational safety interlocks and must not authorize remote actions by itself.
Original practice scenario: A hypothetical equipment analytics team sees inspection measurements shift after a new product mix enters production. How would you investigate calibration drift?
Read model answer +
Separate measurement-system changes from changes in what is being measured. An aggregate shift alone cannot identify whether calibration, process conditions or product composition caused it.
Work through it
Retain instrument and calibration versions, product identifiers, reference measurements and environmental covariates at the observation level.
Compare stable reference samples and like-for-like product slices over time, using a predefined drift statistic and uncertainty rather than searching for the most alarming chart.
Examine missing data and maintenance events, then validate a suspected shift with an independent reference or controlled check before changing downstream correction factors.
Tradeoffs to discuss
Reference checks improve attribution but consume measurement capacity; observational monitoring offers wider coverage while leaving more confounding.
Common pitfalls
Do not normalize away a genuine process change merely to restore a familiar average.
Changing the reference population or instrument software requires a baseline version so monitoring remains interpretable.
