Soft-resolution assessment revision has no accepted all-source Identity revision mapping
Assignment ADR-0053/platform/008, coordinator 01a0e095-f279-7fd0-a8ef-92ce8f9547fc/local. Root alone publishes and routes the exact owner review. This is a source-backed missing-boundary disposition, not an authority amendment or runtime activation.
Accepted Scheduling 3.0 original-unbound-soft-resolution.md lines 45-51 requires GET to remain read-only, a coherent locally locked sidecar/source/fence assessment after fresh external reads, the COMPLETE current Platform verdict inside current_resource, and assessment_revision advancing for every relevant predicate/source fact including staged facts. The adjacent soft-resolution.schema.json defines assessment_revision as a positive safe integer, includes current_resource by the closed Identity schema, and hashes every current-assessment field except the digest itself. A complete digest is integrity evidence, not a monotonic revision source.
Accepted Identity 2.0 scheduling-resource-verdict-api.md defines exactly eight fields. It provides no source revision, event cursor, generation or remote write fence. It explicitly calls evaluated_at observation metadata and accepts a bounded residual identity-change interval after the coherent snapshot. Platform source 1dfa6e7d6ed541e5f7702af905fa4f7919d57c34 implements that closed coherent read and no extra authority. Current person-tenancy/repo.ts upsert changes tenancy status without a published all-source generation. Person update emits person.updated for supported changed fields, but mergePersons changes status/alias and reattaches externals without such an event in this source; Organization existence and all relevant aliases/tenancies also have no single reviewed generation covering every invalidating writer. person.updated or a local Delivery epoch is therefore not an accepted complete source clock.
The narrow interpretation supported by existing text is that observation timestamps may change without a predicate/source-fact change. Identity explicitly makes evaluated_at metadata; Scheduling permits unchanged observations to retain a revision while observed_at/digest change. This supports treating observation metadata separately when determining whether facts changed. It does NOT support excluding person_status, tenancy_status, requested/canonical identity, new_write_eligible, Organization reach or lineage from relevant current predicates. The complete verdict, including evaluated_at, still belongs in the response and digest. Do not silently replace assessment_revision with a timestamp, digest-derived safe integer, receipt revision, writer epoch or local-only mutation counter.
No accepted all-source mapping was found. An unchanged Delivery local assessment revision plus a freshly changed Platform predicate cannot be claimed coherent merely because the complete current_resource is newly hashed. A read cannot increment a persisted revision, consume a sequence or repair a sidecar to conceal this gap. A missing or stale mapping must block positive retirement and preserve original receipts, fences and reconciliation work. This finding does not invalidate the accepted snapshot residual window or demand an exact-instant remote lease; it identifies the distinct meaning and maintenance of Delivery's current assessment revision.
Delivery and Sales should return an exact source/event/revision mapping with normative support, or jointly specify the missing owner assessment maintenance boundary for root approval. A prospective Delivery worker could collect fresh remote facts outside locks, serialize an explicitly recorded local assessment update, then require each read's current non-metadata facts to match that recorded basis or fail closed. That is only a design option for review, not an approved claim that it covers every remote transition or ABA history, and GET still cannot initialize or mutate it. A stronger Platform source-generation design would require a new approved closed wire/version and every invalidating writer's participation. Neither option is implemented or implicitly authorized here.
The separate original-bind evidence gap belongs to Proposed ADR-0054 and does not grant additional authority for this assessment problem. Platform's six ADR-0053 rows and Identity resource candidate remain independently complete local implementation evidence. Owning-domain positive soft-resolution activation and assessment-revision proof remain blocked on this exact boundary; production flags, keys, foreign code and canonical coordination are unchanged.