Module missingness 

Missingness-aware grammar: invalidates drift computation when sensor data is absent for more than MAX_CONSECUTIVE_MISSING_RUNS consecutive runs.

Hardware Reality Check

Semiconductor sensors fail. An MFC with a broken transducer will produce a flat line of zeros, not NaN — and a naive DSFB engine will interpret a sustained zero residual as “nominal” when the truth is “unknown.”

More dangerously, imputed values (mean-fill) can accumulate into a spurious drift signal over a long outage window, causing the engine to escalate a phantom anomaly that exists only in the imputation model.

Policy

• If a feature’s sensor is missing for > MAX_CONSECUTIVE_MISSING_RUNS consecutive runs, the drift value d is invalidated and set to DriftValidity::Unknown.
• Grammar transitions from DriftValidity::Unknown features are suppressed — the feature is held at its last valid grammar state with a suppressed_by_missingness flag set to true.
• The outage event is recorded verbatim in the traceability manifest, preserving the audit trail.

Structs

FeatureMissingnessTracker

Per-feature missingness tracker.

MissingSummary

Compact summary of missingness across all features — emitted in the run manifest JSON for audit.

MissingnessAwareGrammar

Applies the missingness policy across all features in a run sequence.

MissingnessAwareRecord

The annotated record for a single run after missingness processing.

Enums

DriftValidity

Marks whether the first-difference (drift) value for a run is valid.

Constants

MAX_CONSECUTIVE_MISSING_RUNS

Maximum number of consecutive missing runs before drift is invalidated.