Scenario-generator parameters and reproducibility specification

This appendix freezes the v1.2.1 conditional support-pattern generator. The values are author-specified priors, not estimated frequencies, and the outputs are not calibrated forecasts. The generator assumes a qualifying event and complete measurement. It does not simulate event incidence, public observability, null outcomes, mixed outcomes, or the probability of a not-adjudicable verdict. Channel outcomes remain disaggregated.

Component probability ranges and support Boolean patterns are frozen in Table 12.3. Appendix C separately records real-world observability so a clean simulation assumption cannot be mistaken for available evidence. A hypothesis can have a reproducible support-pattern range and still be not adjudicable in practice.

Table C.1. Real-world observability register. Simulation conditions on complete measurement; this table governs the missing-data state.

H	Mandatory evidence	Public status at baseline	Likely evidence owner	Earliest formal endpoint	Missing-data treatment
H1A	Issuer/intermediary spreads, funding, collateral, flows, clearing	Partial	Markets, dealers, funds, clearinghouses, regulators	60 trading days after trigger	Not adjudicable if exposure map or mandatory control is missing
H1B	Shared supplier exposure, issuer operations, index and credit controls	Partial	Issuers, suppliers, exchanges, regulators	120 trading days after trigger	Not adjudicable if common exposure cannot be mapped
H2	Project financing, guarantees, utility/rate, supplier and redeployment data	Limited	Sponsors, lenders, utilities, regulators, municipalities	24 months after trigger	Not adjudicable without financing and replacement maps
H3	Cohort retention, model mix, neutrality, integration cost, ROIC	Mostly private	Acquirer, target, customers, auditors	12 and 24 months post-close	Not adjudicable if mandatory cohort/accounting data remain private
H4	Workload depth, compute-hours, site power, latency, SLA and supply	Private	Operators, customers, utilities	Four quarters after trigger	Not adjudicable without matched site/workload telemetry
H5	Contract, spend, control-plane and substitution panel	Private/collectable	Enterprises, vendors, procurement owners	Annual panel endpoint	Not adjudicable if frozen recruitment or spend denominator fails
H6	Executed failover, RTO/RPO, output quality, TCO	Private/collectable	Enterprises and providers	12 months after shock/test	Not adjudicable without executed tests and costs
H7	Production share, matched task cost, price premium, layer economics	Mostly private	Enterprises, providers, researchers	2030 panel endpoint	Not adjudicable without matched production and spend panel

Four-factor loading rows are ordered market/capital, physical substrate, platform/enterprise, and geopolitics/policy. H1A [0.70, 0.15, 0.10, 0.10]; H1B [0.45, 0.55, 0.10, 0.20]; H2 [0.45, 0.55, 0.10, 0.20]; H3 [0.35, 0.15, 0.60, 0.10]; H4 [0.10, 0.75, 0.25, 0.15]; H5 [0.15, 0.10, 0.75, 0.20]; H6 [0.05, 0.20, 0.65, 0.35]; H7 [0.10, 0.35, 0.55, 0.40]. Hypothesis-specific normal noise supplies the remaining variance. These loadings affect support-component dependence only.

Table C.2. Conditional support-pattern sensitivity cases and fixed execution settings.

Case	Parameter change	Interpretation
Low support prior	Component probability modes x0.80; factor loadings x0.80	Lower conditional support pattern; not a thesis-null or missing-data scenario.
Base	Frozen Table 12.3 priors, seed 20260724, 200,000 trials, correlation 0.35	Reference conditional support-pattern generator.
High support prior	Component probability modes x1.20; factor loadings x1.20 with residual variance positive	Higher conditional support pattern; not an observed probability.
Sensitivity envelope	400 batches x 4,000 trials; correlation 0.10-0.60; triangular low/mode/high priors	Produces 5th-95th support bands only.

Reproduction sequence: (1) draw four standard-normal factors and hypothesis-specific idiosyncratic errors; (2) construct correlated latent variables from the frozen loading rows; (3) transform to component uniforms with the normal CDF; (4) compare each component with its triangularly sampled probability; (5) apply the hypothesis support Boolean; (6) repeat for base and sensitivity cases; and (7) report conditional support frequency only. Real-world event occurrence, observability, null, mixed and not-adjudicable states are handled by Tables 12.1, 12.1A, 12.2 and C.1. Later parameter changes require a prospective versioned amendment and do not overwrite v1.2.1.

APPENDIX D