D13's decision table is short one input class — carry the correction at a milestone

[guycorbaz]

The gap, measured

architecture.md's D13 gives the verdict algebra as a six-row table over the verdict set (architecture.md:967-974). Enumerated over the presence of each verdict, that table leaves exactly one input class with no row:

≥1 Opposes, with no Decisive, no Supports and no Disqualifying.

It is not "only Neutral / nothing" (there IS an Opposes), it is not "Supports AND Opposes" (there is no Supports), and every remaining row requires a Decisive.

The eight combinations of (Decisive, Supports, Opposes) that survive Disqualifying = absent:

Decisive	Supports	Opposes	D13 row
✔	✔	✔	:971
✔	✔	✗	:970
✔	✗	✔	:971
✔	✗	✗	:970
✗	✔	✔	:973
✗	✔	✗	:972
✗	✗	✔	NONE
✗	✗	✗	:974

Re-derived independently five times: at story 5.4's contexting, by both of its fresh-context validation agents, by its code review's Edge Case Hunter (16-way enumeration), and again at story 5.4b's contexting.

The arbitration, already taken and already implemented

Guy's call, 2026-07-29: that class concludes Abstained { AbsenceOfProof }. Nothing argues FOR the merge, so there is no merge to refuse, and D13 deliberately reserves the refusal-that-names-a-rule for Disqualifying.

Story 5.4b implements it in crates/opencmdb-core/src/identity/cascade.rs's decide, in an arm explicitly labelled as the arbitration and citing no architecture line — because there is none to cite. A match on the presence tuple makes a missing class an error[E0004] rather than a silent fallthrough (measured: deleting the arm gives error[E0004]: non-exhaustive patterns: (false, false, false, true) not covered).

What this issue is for

The correction to D13's own table body in architecture.md is a MILESTONE edit, never a story task — epics.md:1461 says so in the acceptance criterion that created story 5.4b. A story that patched a locked decision document would be exactly the drift the register exists to prevent.

So this issue carries the promise. At the next milestone:

• add the seventh row to D13's table at architecture.md:967-974, or restate the existing rows so the class is covered
• record the arbitration and its date in the decision body
• regenerate architecture-views.md in the same pass (its sourceSha256 is already stale — issue architecture-views.md is STALE: regenerate at the next milestone (drift measured: 2 commits, 59 insertions) #50)

Until then the authority is the code and its doc: decide's rustdoc carries the full enumeration so a reader can re-derive the gap rather than trust it.

Related

• architecture-views.md is STALE: regenerate at the next milestone (drift measured: 2 commits, 59 insertions) #50 — architecture-views.md staleness, the other milestone-owned documentation debt
• Story 5.4b (_bmad-output/implementation-artifacts/5-4b-verdict-algebra-and-float-gate.md), AC3