The machine-readable model is COVERAGE_MODEL.json. Coverage combines:
- exhaustive critical protocol × family × action interactions;
- explicit stateful issue #206/#209, lifecycle, UDP and security scenarios;
- deterministic pairwise coverage for payload, close mode, selector, endpoint, concurrency, topology and timing;
- disabled application/performance placeholders;
- explicit unsupported product-scope declarations.
The catalog is a declaration inventory, not a count of working tests. Reports
separately count declared, executable, declarative, unsupported and selected
entries. They distinguish DRY_RUN_READY, NOT_IMPLEMENTED, MOCK_PASS,
MOCK_EXPECTED_FAIL, MOCK_HOLD, TESTED_PASS, TESTED_FAIL, EXPECTED_FAIL,
BLOCKED_BY_KNOWN_DEFECT, SKIPPED_CAPABILITY, SKIPPED_SELECTION,
and UNSUPPORTED_PRODUCT_SCOPE. Literal all-packet
exhaustiveness is impossible and is not claimed.
TESTED_PASS and TESTED_FAIL are emitted only from real runtime. Dry-run and
mock coverage retain their own statuses and never become tested product claims.