Validation: run one independently authored substrate/adapter round against the world contracts

[zachshallbetter]

Parent: #1111
Depends on: #1113, #1114

Goal

Address the largest evidence weakness in the current world-substrate program: most theory, substrate selection, adaptation, and grading have been performed within one authorship context.

Run at least one adversarial-reuse round where the substrate and/or adapter is authored independently of the person who defined the FE abstraction being tested.

Minimum independence target

Prefer, in order:

1. external substrate + external adapter author;
2. external substrate + independent adapter author;
3. external substrate + FE author adapts only after predictions are frozen.

Do not count a synthetic fixture written specifically to exercise FE as independent evidence.

Protocol

• select a materially different, already-existing system with its own native semantics;
• freeze the source/version before FE mapping;
• preregister predicted contract fit, expected churn, and likely failure points;
• do not expose the adapter author to desired outcomes where avoidable;
• record every requested contract change and classify structural / representational / convenience;
• preserve abandoned mappings, rejected changes, and surprises;
• have an independent reviewer grade semantic distortion if possible.

Evidence outputs

• source provenance + version;
• prediction record;
• adapter diff;
• contract-churn report;
• semantic-loss/distortion report;
• negative controls;
• qualification record under Governance: put world-substrate research under IEPE/EQP evidence + promotion discipline #1114.

Acceptance

• One complete round exists with provenance stronger than the current self-authored substrate evidence.
• The outcome may be generalizes, generalizes-with-refinement, field-fitted, or inapplicable; none is privileged.
• No contract change is accepted merely to make the experiment pass.
• The resulting evidence is usable by Research method: boundary extraction under adversarial reuse #1113 when evaluating whether “boundary extraction under adversarial reuse” is more than an internal design habit.