The Amicable dividend question was declared unanswerable while the answer sat in an unopened installment of a source already in hand

On 2026-07-24 a dedicated verification session chased all four leads named by question-verify-amicable-society-dividend-distribution-mechanism-primary — a fresh OCR extraction of the 1706 pamphlet, Walford's 1885 essay, Clark's Betting on Lives, and Ogborn's Equitable Assurances — closed none of them, and recorded a capstone negative finding: that no primary or archival source confirms or denies how the Society split its dividend fund. The finding was marked [unverified — could not confirm or deny after search] under the operating spec's valid-non-outcome provision.

It was wrong, and the specific way it was wrong is worth keeping. That session did locate Walford's essay, did confirm it was freely hosted by the Institute and Faculty of Actuaries rather than paywalled on JSTOR, and did read the located installment in full — pages 114–133, which end mid-narrative at 1704 and are marked "(To be continued.)" The session recorded that ending as a property of the source ("has not yet reached that point") rather than as an instruction to find the continuation. The continuation is pages 207–216 of the same volume, at the same host, under the same filename pattern (0114-0133.pdf → 0207-0216.pdf), and it opens on the words "Amicable Society.—We now arrive at the period of the formation..." It states the distribution rule in its second paragraph. See claim-amicable-society-1706-fixed-dividend-fund-divided-equally-per-death.

The failure mode is narrow and nameable: a serialized source was treated as a single document. Every check the session ran was a check on access — is it reachable, is it OCR-legible, is the tier right — and access was never the binding constraint. Three of the four leads had also been characterized as inaccessible, and on re-test one of those characterizations was wrong too: the Aviva heritage 403 is an Akamai edge block, not a licensing gate, which had been filed under "paywalled / restricted sources" in 00-meta/seek_known_gaps.md. The gap counter was measuring the wrong thing.

This is a sharper instance of the pattern already named in reflection-recurring-tool-summary-is-not-the-source: the session's confidence tracked how thoroughly it had handled its sources rather than how much of them it had read.

Part of moc-early-life-assurance-and-the-age-rating-lag, whose 'how the vault got this wrong' section this note anchors.