Get this run's structured results (poll-able, per-document)

The same governed results read as GET /v1/pipelines/{id}/results (review holdback, demoted-field projection, assembly awareness), scoped ALWAYS to this /v1/run request's own documents[] echo and ALWAYS the per-document view — "what did MY submitted documents produce," uniform across new/append pipeline modes. This is a deliberate DIVERGENCE from the per-run run.completed webhook, which for an unscoped non-append assembly run delivers COMPOSED rows: composed output for an assembly run is GET /v1/pipelines/{id}/results?view=composed, surfaced here at links.composed_results when a pipeline exists.

Only document_id and include filter the rows; run_id, view, since, until, and status are pipeline-route-only and 400 with unsupported_filter here, since the run IS the scope and its window is already bounded. include=cells,provenance,audit behave identically to the pipeline route — including the structured-subfield type projection on fields/cells and the raw, non-projected audit trail — see GET /v1/pipelines/{id}/results for the full shape and held-value redaction rules.

Two zero-states return 200 with an empty page and a scope_note, never 404: no compiled pipeline exists yet for this run (still ingesting, or it failed before pipeline creation), or the run predates per-request document tracking (an echo-less legacy row, whose membership is unknowable). The envelope's status folds the same three-state vocabulary as GET /v1/run/{id} (processing/completed/failed) and DOES settle even for an append pipeline that never settles at the pipeline level — poll this route for append consumers instead of the pipeline route. Requires read scope.