docs(audit): close COH-009 — BR-003 conflation already split into BR-010

[beyondnetPeru]

COH-009 asks for a single business-rule id carrying two rules to be split.

The split already exists, and the business-rule table credits this row for it. docs/diagrams/evolith-tracker-crosscutting-diagram.md:163 carries a BR-010 row, CFR Quality Threshold, whose own description reads «escindida de BR-003 por COH-009; BR-003 es sólo la firma humana».

Not taken on that line's word — verified across the repository:

• every BR-003 occurrence outside docs/audit/ means Human Sign-Off and nothing else;
• the QA-gate half is cited as BR-010 by reference/specs/qa/README.md («no deployment is authorized without an approved QA verdict and CFR < 2%»), by reference/specs/testing/tracker-test-strategy.md against PhaseGateEvaluator, and by reference/specs/design/tracker-redo-flow-design.md.

Two rules, two ids. The work was done and the row was never moved.

This is the eighth row found this week describing work already finished. Its claim is about meaning, not presence, so it is deliberately not added to falsifiable-claims.json — that file takes only what a probe can refute without judgement, and a regex over prose is precisely the failure mode it exists to avoid.

Board: 6 OPEN, 1 DEFERRED, 100 RESOLVED (107 ids before and after).

🤖 Generated with Claude Code

[beyondnetPeru]

Found after opening this PR, and it strengthens the case while exposing a second problem:

docs/audit/tracker-gap-tracking.md:130 already had COH-009 as DONE, with the same resolution stated: «Escindido: BR-003 conserva Human Sign-Off (canónico); el umbral CFR<2% pasa a nuevo BR-010, referencias actualizadas EN+ES».

So the Tracker carries two gap boards that disagreed about this row — one DONE, one OPEN — and neither is derived from the other. That is the same class of defect COH-009 itself describes: one thing tracked under two identities, where satisfying one reads as satisfying both. Filing it as its own row rather than smuggling it into this one.