The facets container derives from the store rather than minting at capture

@asFacet is the one expansion the capture walk cannot run, and the fact store currently has no other home for it, so the <Conn>Facets and <Scalar>FacetValue shapes exist only in the assembled-schema synthesis. Capture expands a macro when its contribution is a function of one carrier’s own declaration, which is what keeps a single file the unit of an incremental refresh. The facets container fails that test: its shape reads through the carrier’s arguments into the filter input type’s fields, for the @asFacet marker, the @field(name:) binding, and the value’s scalar and nullability. That input type is free to live in another file, so minting the container during the walk would leave it stale whenever a facet is added to a file the carrier’s own refresh never re-reads. It is an aggregate over the whole schema, not a local expansion.

Every input the shape needs is already a captured column, which is what makes this a derivation rather than a gap in capture: graphitron_facet marks the input fields, graphitron_field_binding gives each one’s bound column, graphql_argument links a carrier to its filter type, and graphql_field carries the value’s scalar and nullability. The container therefore computes as a query over captured columns, which is exactly the boundary the fact-base’s decode rule draws when it keeps name resolution and effective-value defaulting out of the walk.

Open questions for the Spec pass. Where the derived types live, given that the base relations hold declarations and a derived type is not a declaration anyone wrote. Whether the facets field on a Connection is derived alongside the container or minted with the rest of the Connection’s fields and left dangling until the derivation runs. Whether the first-occurrence-wins dedup on a duplicate facet name across a carrier’s filter inputs stays where ConnectionPromoter.facetSpecsFor has it, or becomes a detection now that the store can express the duplicate. And what the derived stratum’s provenance looks like, since the macro domains on the capture-side provenance relations deliberately carry no FACET value.