diff --git a/docs/current-handoff.md b/docs/current-handoff.md
index 6de96b0..8d959dc 100644
--- a/docs/current-handoff.md
+++ b/docs/current-handoff.md
@@ -145,6 +145,12 @@
 - OpenAI strict transport projection now derives structure from canonical Zod-generated JSON Schema and canonical input optionality from the Zod contract, including `.optional().default(...)`. Canonically omittable fields are required-but-nullable for transport; null placeholders are omitted only for those fields before canonical Zod validation, while genuinely required fields remain non-nullable.
 - `reconstructionV2Schema` and Ollama behavior remain unchanged. `START_CASE_EXPERIMENT_PROVIDER=openai` is experiment-only selection; default behavior remains Ollama/Qwen. Deterministic closeout passed with zero live calls; next boundary is exactly one live GPT-5.6 Terra reconstruction.
 
+## Reconstruction-only OpenAI observation seam
+
+- The first Terra observation was blocked before execution because normal `startCase()` would continue into Ollama-backed question generation. No Terra inference occurred.
+- An experiment-only reconstruction-only seam now returns after initial reconstruction and graph construction, before `determineGraphBackedQuestion`. Normal production/default `startCase()` behavior remains unchanged; explicit OpenAI helper selection guarantees one selected reconstruction provider call and zero downstream question-generation calls.
+- Zero live calls occurred while implementing and verifying this seam. Next boundary: exactly one live Terra reconstruction with no retry.
+
 ## Current product architecture
 
 Three distinct routes, not a single page:
diff --git a/lib/graph/orchestrator.js b/lib/graph/orchestrator.js
index fcc725b..1604d0f 100644
--- a/lib/graph/orchestrator.js
+++ b/lib/graph/orchestrator.js
@@ -412,6 +412,26 @@ export async function startCase(body, dependencies = {}) {
   const graphReferenceValidation = validateGraphReferences(
     initialSituationGraph,
   );
+
+  if (dependencies.reconstructionOnly) {
+    return {
+      success: graphReferenceValidation.valid,
+      summary: analysis.reconstruction?.summary ?? null,
+      reconstruction: analysis.reconstruction,
+      situationGraph: initialSituationGraph,
+      selectedQuestion: null,
+      diagnostics: buildDiagnostics({
+        analysis,
+        graph: initialSituationGraph,
+        graphReferenceValidation,
+      }),
+      validationErrors: graphReferenceValidation.valid
+        ? undefined
+        : graphReferenceValidation.errors,
+      statusCode: graphReferenceValidation.valid ? 200 : 500,
+    };
+  }
+
   const provider = dependencies.provider ?? getProvider();
   const modelName = dependencies.modelName ?? analysis?.modelName ?? null;
   const initialQuestionResult = await determineGraphBackedQuestion({
diff --git a/scripts/start-case-experiment-helper.cjs b/scripts/start-case-experiment-helper.cjs
index 82556df..6871dbf 100755
--- a/scripts/start-case-experiment-helper.cjs
+++ b/scripts/start-case-experiment-helper.cjs
@@ -64,6 +64,7 @@ async function runStartCaseExperiment(scenarioInput, experimentInstruction, opti
   let startCase = options.startCase;
   let reconstructionProvider = options.reconstructionProvider;
   let reconstructionModelName = options.reconstructionModelName;
+  let reconstructionOnly = options.reconstructionOnly;
   const isMock = process.env.START_CASE_EXPERIMENT_HELPER_MOCK === "1";
 
   if (process.env.START_CASE_EXPERIMENT_PROVIDER === "openai") {
@@ -80,6 +81,7 @@ async function runStartCaseExperiment(scenarioInput, experimentInstruction, opti
       fetchImpl: fetch,
     });
     reconstructionModelName = "gpt-5.6-terra";
+    reconstructionOnly = true;
   }
 
   if (!startCase && isMock) {
@@ -120,6 +122,7 @@ async function runStartCaseExperiment(scenarioInput, experimentInstruction, opti
     result = await startCase(scenarioInput, {
       reconstructionProvider,
       reconstructionModelName,
+      reconstructionOnly,
     });
   } finally {
     // Clean up experiment env var after execution regardless of outcome
@@ -141,6 +144,7 @@ async function runStartCaseExperiment(scenarioInput, experimentInstruction, opti
   return {
     success: result.success,
     summary: result.summary ?? null,
+    reconstruction: result.reconstruction ?? null,
     situationGraphNodeCount: result.situationGraph?.nodes?.length ?? 0,
     situationGraphEdgeCount: result.situationGraph?.edges?.length ?? 0,
     selectedQuestion: result.selectedQuestion?.question ?? null,
@@ -148,6 +152,7 @@ async function runStartCaseExperiment(scenarioInput, experimentInstruction, opti
     assessmentProgress: result.assessment?.progress ?? null,
     endToEndElapsedMs: Date.now() - endToEndStartedAt,
     diagnostics: result.diagnostics ?? null,
+    providerApiPath: result.providerApiPath ?? null,
     error: result.success ? null : (result.error ?? "unknown"),
     statusCode: result.statusCode ?? (result.success ? 200 : 500),
   };
diff --git a/tests/graph/orchestrator.test.js b/tests/graph/orchestrator.test.js
index 4fe3f4f..a4728c3 100644
--- a/tests/graph/orchestrator.test.js
+++ b/tests/graph/orchestrator.test.js
@@ -374,6 +374,34 @@ describe("lib/graph/orchestrator startCase", () => {
     });
   });
 
+  it("returns initial reconstruction evidence before downstream question generation", async () => {
+    const analysis = makeAnalysisResult();
+    mockAnalyseScenario.mockResolvedValue(analysis);
+    const downstreamProvider = {
+      generateReconstruction: vi.fn(() => {
+        throw new Error("downstream question generation must not run");
+      }),
+    };
+    const { startCase } = await import("@/lib/graph/orchestrator.js");
+
+    const result = await startCase(
+      { scenario: "Scenario text" },
+      { reconstructionOnly: true, provider: downstreamProvider },
+    );
+
+    expect(result).toMatchObject({
+      success: true,
+      reconstruction: analysis.reconstruction,
+      selectedQuestion: null,
+    });
+    expect(result.situationGraph.nodes.length).toBeGreaterThan(0);
+    expect(result.diagnostics.graphReferenceValidation).toEqual({
+      valid: true,
+      errors: [],
+    });
+    expect(downstreamProvider.generateReconstruction).not.toHaveBeenCalled();
+  });
+
   it("rejects invalid request input without throwing", async () => {
     const { startCase } = await import("@/lib/graph/orchestrator.js");
 
diff --git a/tests/scripts/start-case-experiment-helper.test.js b/tests/scripts/start-case-experiment-helper.test.js
index 21d934a..b382ecd 100644
--- a/tests/scripts/start-case-experiment-helper.test.js
+++ b/tests/scripts/start-case-experiment-helper.test.js
@@ -174,6 +174,7 @@ describe("start-case-experiment-helper.cjs apparatus", () => {
             startCase: async (_input, dependencies) => {
               expect(typeof dependencies.reconstructionProvider?.generateReconstruction).toBe("function");
               expect(dependencies.reconstructionModelName).toBe("gpt-5.6-terra");
+              expect(dependencies.reconstructionOnly).toBe(true);
               return {
                 success: true,
                 situationGraph: { nodes: [], edges: [] },