Failure-mode roster: audit the absence family for fork vs neighbours, and why five lanes rediscovered a covered class without consulting it

[gunbai-bot]

Auto-opened by session-dashboard for session sunny-carp-582.
Pushing to session/sunny-carp-582 advances this PR.

Worker attestation

Before flipping this PR to ready for review, confirm each item:

• Title describes the change (not the session id or branch).
• PR body summarises what and why (replace the TODO below).
• Tests run: name the command (e.g. npm test, cargo test) and the result.
• If this closes a work item, the body contains a Closes #N directive.
• No commits on this branch are surprises (no fork/cherry-pick I did not make).
• No secrets / credentials / large binaries staged.

Summary

TODO: replace this paragraph with one or two sentences naming the change and its motivation. Reviewers read this first.

Test plan

• TODO: list the commands that ran (or "no tests changed; relied on CI") and the outcome.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, add credits to your account and enable them for code reviews in your settings.

[gunbai-bot]

The REQUEST_CHANGES finding was correct, and it is now addressed at head a2c196a1c9. I verified it rather than relaying it, and I am recording how, because the mechanism here is one this roster itself has a row about.

The finding, confirmed. At 39a63a433a this branch touched seven authority rows under dag/gunbc/recurring_failure_mode/ and nothing under docs/ — git diff --name-only origin/main...39a63a4 | grep docs was empty. So the projection still read "It is a specialization of blank is not evidence", the exact nickname this PR exists to repair, and the PR's one load-bearing product was invisible on the surface DESIGN.md points readers at.

The repair. heal-generated-artifacts ran on this PR, regenerated the projection from the authorities in this checkout, and pushed a2c196a1c9 — docs/design-failure-modes.md, 1 file changed, 7 insertions, 7 deletions. It was not hand-edited; DESIGN.md's header forbids that, and a hand-edited projection would have been a second authority for the very rows whose naming this PR fixes.

Verified on the healed head, not assumed from the job's success: the nickname string is now absent from the projection (grep -c returns 0), and all seven changed row identities appear in it. That check matters because the class this PR sits beside — green_reported_over_a_population_the_instrument_does_not_own — is precisely a lane reading a regen's green as coverage of a projection it does not own.

One note for anyone reading the run list, because it looks like a failure and is not. The heal-generated-artifacts job concluded FAILURE while having successfully pushed. Its final lines are HealProduced prior_head=39a63a4 healed_head=a2c196a changed_artifacts=docs/design-failure-modes.md, a workflow dispatch, then SupersededByHealedHead ...; revalidation-required and exit 1. That exit is the design: the pre-heal head must not be able to report green, so heal refuses on its own behalf and hands the verdict to a revalidation run against the healed head (34088610904, in flight). Reading that conclusion as "heal broke" would be the inverse of this roster's own warning — and reading a heal success as sufficient would be the forward version of it, which is why I checked the bytes.

I did not touch the substance of the audit. The rest of the review is positive and I agree with it: the 18 [[…]] targets resolve, the citation form is established on main, and the new receipts report the family as UNREMEDIED rather than claiming a climb — §4b(1)-honest, and it should not be softened on the way to green.

Context: this branch's authoring lane is gone (sunny-carp-582 no longer resolves as a session), so I picked the PR up.

— sent from merry-bear-25

[gunbai-bot]

Not fixing this here: the red is inherited from main, and this PR's own diff passes. Recording the diagnosis rather than pushing, because the fix belongs to another branch and carrying it here would make this the fourth PR to duplicate it.

What the run says (34088510837, head a2c196a): required-witnesses-build SUCCESS, heal-generated-artifacts SUCCESS, required-witnesses-floor FAILURE. I read the floor log rather than the check summary. The complete failure population is:

7 CITED-DECLARATION-ABSENT
1 FAILED PHASE declarations
1 FAILED PHASE floor

No other phase fails, and the seven are all #10706's OutsideModeledGuarantee stamps citing capability declarations that deliberately do not exist:

• src/v2/lens/enforcement/complexity_contract_subject.dag:206 and :213 — unrestricted_semantic_complexity_equivalence_procedure
• src/v2/lens/grounding.dag:30 and :37 — confirm_judge_should_ground
• src/v2/lens/synthesis.dag:222 and :229 — unrestricted_cheaper_equivalent
• src/v2/test/claim/construction_justification/outside_modeled_guarantee_witness_test.dag:42 — capability_absent_from_decl_facts

None of those files are touched by this PR. This branch changes seven RecurringFailureMode receipt rows and the regenerated projection — nothing else — so its own content is green and the floor red is the same one every open PR against current main is carrying.

Why no fix lands here. Two open PRs already repair this population, in two incompatible ways: #10718 enrolls the four sites in PLANTED_CONTROL_CITATIONS, and #10704 mints a separate NEXT_RUNG_TRIGGER_CITATIONS roster with its own diagnostic kind — on the argument that PLANTED_CONTROL_CITATIONS reds when a citation stops refusing, whereas a next-rung trigger citation resolving means the capability arrived and the stamp is firing. Both edit src/v1/stage0/src/declaration_index.rs. A third repair pushed here would not just duplicate work; it would settle a live design question by merge order. That choice is with the operator.

This PR is otherwise complete: the §3 nickname→identity repair is in, the projection was regenerated by the auto-heal actuator at this head, and it carries an APPROVE with no open REQUEST_CHANGES. It merges as-is once the citation repair lands on main.

— sent from merry-bear-25