Document retry composition and preserve error summaries

docs/operations.md

@@ -163,6 +163,17 @@ operator needs one HTTP attempt per generation. A semantic correction rebuilds
 the ordinary producer request and supplies only the latest rejected model
 response plus aggregated validator guidance; it is not a conversation replay.
 
+These budgets multiply only by their declared scopes. With two outer producer
+retries, two transient retries, and one structural repair, a single producer
+can issue up to 18 HTTP attempts: three producer attempts, each with an initial
+generation and one repair generation, each allowing three HTTP attempts. This
+is an illustration, not a total-deadline or exactly-once guarantee. Every
+built-in HTTP attempt has its own generation timeout; caller cancellation
+remains the overall authority and interrupts Retry-After waiting. The shared
+Notarius scheduler keeps its permit through those waits, so retry budgets can
+increase latency and provider cost. Observed usage can remain incomplete when
+providers omit accounting fields.
+
 After the applicable budgets are exhausted, the resolved
 [`validation_policy`](config.md#pipelines) determines the result. Structural
 failure and semantic rejection normally fail the run; an explicit

docs/roadmap/implementation.md

@@ -310,6 +310,8 @@ the new projection assertions. No new warnings result from recovered retries.
 
 ## Stage 7 — Verify Retry Composition And Finish Operator Guidance
 
+**Status:** Completed 2026-09-25.
+
 **Goal:** Validate the cross-boundary behavior that individual stages cannot
 prove, then reconcile current documentation. One prompt; no generalized retry
 test framework or runtime redesign.

internal/framework/llm/scheduled_client.go

@@ -38,7 +38,7 @@ func (c *scheduledClient) CompleteStructured(ctx context.Context, req contracts.
 		return callErr
 	})
 	if err != nil {
-		return contracts.StructuredCompletionResponse{}, err
+		return response, err
 	}
 	return response, nil
 }