The provenance chapter says the source-hash registry is durable and only partly regenerable; itd-36 scopes it as regenerable.

The provenance chapter says the source-hash registry is durable and only partly regenerable; itd-36 scopes it as regenerable.

A premise contradiction (severity minor) found by the cross-document consistency pass.

• End A: .abcd/development/brief/05-internals/09-provenance-substrate.md:52 — "DURABLE source metadata, only PARTIALLY regenerable"
• End B: .abcd/development/intents/shipped/itd-36-memory-unification.md:48 — "source-hash registry at .abcd/memory/.sources_index.json (regenerable per the lifecycle taxonomy)"

09-provenance-substrate classifies .abcd/memory/.sources_index.json as durable source metadata that is only partly regenerable. It says fields about a discarded original cannot be rebuilt by a crawl, and that this supersedes the blanket 'regenerable' classification. itd-36's scope still files the same registry as regenerable under the lifecycle taxonomy. A builder following itd-36 could drop and rebuild a file that the substrate says it cannot rebuild.

Evidence: .abcd/work/reviews/2026-10-26-consistency/00-summary.md, finding 1024.

Grounds

• declined: Closed in the 2026-10-27 triage (dcn-2610271112330536): a later record already settles which end holds. The substrate chapter cites the ADR and states the supersession explicitly; the shipped intent's scope line was true before that refinement.