evidence(OMN-15597): OCC companion for OmniNode-ai/omnimarket#1998

[onexbot-occ-writer]

Deterministic OCC evidence companion for OmniNode-ai/omnimarket#1998, authored by node_occ_companion_effect (RSD-3, OMN-14622) from the node_occ_companion_compute plan. Every byte is a pure function of the compute plan (attestation-oracle reproducible).

[jonahgabriel]

FLAG (no action taken) — stale second companion for an already-merged PR

Raised by lane fable-ws0-0731/laneG4 as an observation for the sweep. This lane has taken no action on this PR — not closed, not merged, not rebased.

Facts, live at 2026-08-01:

• This PR is an OCC evidence companion for fix(OMN-15597): un-vacuate the AC5 census oracle and pin AC5 to a committed OCC corpus omnimarket#1998.
• evidence(OMN-12473): bind node_model_eval_orchestrator proof #1998 is already MERGED — 2026-08-01T11:51:46Z, merge commit 3cb2633c06b337eb19d3a31840304e9b654b0526, now on omnimarket@dev.
• The binding companion for evidence(OMN-12473): bind node_model_eval_orchestrator proof #1998 was evidence(OMN-15597): OCC companion for OmniNode-ai/omnimarket#1998 #5827, merged 2026-08-01T11:09:48Z — 42 minutes before evidence(OMN-12473): bind node_model_eval_orchestrator proof #1998 merged. That is the one evidence(OMN-12473): bind node_model_eval_orchestrator proof #1998's body cites as Evidence-Source: OCC#5832… which is itself a mismatch worth noting: the body cites this PR while the companion that actually satisfied occ-preflight / eligibility at merge time was evidence(OMN-15597): OCC companion for OmniNode-ai/omnimarket#1998 #5827.
• This PR was created 2026-08-01T11:31:58Z, i.e. after evidence(OMN-15597): OCC companion for OmniNode-ai/omnimarket#1998 #5827 had already merged, making it the second companion for the same source PR. mergeStateStatus is currently BLOCKED.

So it is a duplicate-companion residual, not a missing-evidence problem: eligibility for #1998 was satisfied at merge time by #5827. Disposition (merge as an additional observation append vs. close as superseded) belongs to the Codex merge sweep, not to a build lane.

Related, so the sweep has the full picture: #1998's content carried a false-provenance defect (the AC5 corpus was pinned at onex_change_control@9089ffa9, which is not OMN-15597's census point — that is d7b16271). The repair is omnimarket#1999, and a dated correction block has been prepended to #1998's body pointing at it. #1999 will mint its own companion; this PR is not it.

Context: OMN-15597, Linear comments 7d7e31ff (#1998 adversarial verifier) and 921b6c7b (lane G3).

[jonahgabriel]

Closing as superseded/unneeded. OCC#5827 merged first (04a343c52e26b96583626f097952595f8fcc6754) as the earlier companion for omnimarket#1998, but it did not bind #1998’s final 152d62337bc638b585f85f9927975ae82043cab7 head. #5832 was the attempted later re-mint for that final head, but it never merged and fails append-only; it is not a valid Evidence-Source. The underlying false-provenance defect was repaired by merged omnimarket#1999 (58de0731d6398508e43e12ad1d2ad939fc29028a) with merged OCC#5845 (51e0dc7695f5f5ebddee945382ccb7d5bfd5ef54). Therefore #5832 is not the successor evidence and must not merge. #1998’s body is being corrected to an explicit historical/supersession note, not rewritten to falsely claim #5827 exact-final-head coverage.