G.11:4.3a - Causal-use refresh sentinels
When a shipped result consumes C.28, refresh planning watches the causes that can change a supported use, unsupported use, support-result verdict, limits, or downstream decision basis:
| Sentinel | Affected result | Refresh pins |
|---|---|---|
| sampling-realizability shift | CounterfactualSamplingRealizabilityResult | target distribution, decision Method and any derivation, physical, ethical, operational, and history constraints, required sampling construction or obstruction, unresolved questions, status, supported use, and unsupported use |
| performed sampling or resulting-data shift | dated sampling Work plus A.10 evidence path and empirical data regime | WorkPlan when used; actual performer identified through A.13; dated Work independently admitted through A.15.1; Method and window; resulting sample or data; provenance and currentness. Add F.6 with the same A.13 assignment only if the refresh decision needs to say exactly under which assignment the Work was performed. F.6 identifies neither performer nor assignment; a missing or failed attribution leaves the Work intact. A realizability result cannot substitute. |
| identification or bound shift | CausalIdentificationResult | data-regime refs, assumptions, identifying derivation, bound or failure witness, sensitivity |
| estimate shift | CausalEstimateResult | identification or design basis, data, estimator Method, diagnostics, uncertainty, sensitivity, and any live estimation-consistency result |
| target-trial practice shift | protocol and mapping results | question/estimand, protocol-to-data mapping, assumptions, estimate/precision, sensitivity and reporting source edition |
| causal-fairness shift | C.28 support result plus D.5 BiasAuditReport@Context | fairness estimand, extra counterfactual-identification assumptions, estimate and consistency result when used, support components and result, affected population, audit limits, and decision |
| causal-representation shift | CausalVariableRepresentationRecord | intervention validity, invariance, abstraction fidelity, query preservation, shift and use limits |
| off-policy or causal-RL shift | OffPolicyCausalEvaluationResult | behaviour/evaluation policies, horizon/history, confounding, overlap, endpoints, estimator and uncertainty |
| simulation-validation shift | simulationResultRef in CausalSupportComponentRefs | model assumptions, validation, sensitivity, supported model use and unsupported realized/interventional use |
| transport-endpoint shift | CausalTransportabilityResult | source/target population, domain, environment and data-generating regime, assumptions, windows, formula and unresolved limits |
These are payload distinctions under existing G.Core trigger kinds, not new trigger kinds. Reopen only the affected result and downstream uses that consumed it.