Evaluate design against an unrelated-lineage test corpus before freezing

[lann]

Both systems the design has been checked against share one lineage (webcrypto's conformance system forked from webrtc-datachannels' suite), and this contract was synthesized from webcrypto's — so within-family agreement is weak evidence (inheritance, possibly inertia). Independent support so far: external framework precedent (pytest/libtest-mimic/JUnit/NUnit/Go conventions), CM ABI constraints, prototype findings.

Before 0.1.0 freezes the remaining surfaces (L1 is close; L4 schema next), evaluate against at least one corpus with no shared ancestry. Candidates:

• WPT: harness conventions, testharness.js statuses (incl. its timeout/notrun vocabulary), expectation files at Chromium/Firefox scale — stresses the L4 vocabulary and ratchet design.
• go test: the test2json protocol as a results-format precedent; subtests (t.Run) stress the flat-case-list assumption; t.Parallel stresses the concurrency declaration model.
• LLVM lit: substitution/parameterization model stresses ids-encode-parameters; REQUIRES/UNSUPPORTED/XFAIL directives are a mature, unrelated feature-mark precedent (closest external analog to /! marks).

Output: a gap list in the same form as the webrtc evaluation, before the results-schema design starts.

[lann]

Evaluation completed (WPT, go test/test2json, LLVM lit). Verdict: flat slash-named case identity, stream→document duality, structured per-case diagnostics, and unknown-status tolerance all validated by independent lineages; consensus gap found and adopted (not-reached status + unterminated-segment fold rule); provenance-matchable ratchets and per-target ratchet keying adopted; intermittent-sets and retry-visibility reserved as extension points; the skipped-rename suggestion was examined and declined (go t.Skip and pytest runtime-skip share our run-then-bail semantics — the 'all three agree' consensus was 1.5 of 3). Full deltas incorporated into the L4 schema design issue.