fix(confidence-engine): scope contribution filtering to targetNodeId in composeDeepInvestigationCheckpoint

Historical contributions for a different Open Question node no longer
suppress fresh focused deconstruction of the current target. Only
contributions whose targetNodeId matches the current target are counted
towards suppressing redundant deconstruction.

tests/graph/live-update-experiment-helper-checkpoint.test.cjs

@@ -462,6 +462,22 @@ assert(typeof composeDeepInvestigationCheckpoint === "function", "composeDeepInv
 // AC3. The helper structure is intact — no production code was rewritten
 assert(helperSrc.includes("const provider = getProvider()") && helperSrc.includes("const timedProvider = {"), "helper still creates both raw provider and timedWrapper (architecture preserved)");
 
+// ---- ITERATION 15: TARGET-SCOPED CONTRIBUTION HANDLING ──────────────────────
+// Historical contributions for node A must not suppress deconstruction for node B.
+console.log("\n=== ITERATION 15: TARGET-SCOPED CONTRIBUTION HANDLING ===\n");
+
+// Case A: Empty contributions trigger fresh deconstruction — verified by condition using effectiveContributionsForTarget.length (0) when no matching contributions exist for the target
+assert(helperSrc.includes("effectiveContributionsForTarget.length") && helperSrc.includes("=== 0"), "Case A: deconstruction guard checks effectiveContributionsForTarget length (empty → trigger deconstruct)");
+
+// Case B: Contributions filtered by targetNodeId — historical contributions from a different node must not block fresh deconstruction of the current target
+assert(helperSrc.includes("effectiveContributions.filter") && helperSrc.includes("targetNodeId"), "Case B: contributions filtered by targetNodeId before length check (scoping mechanism present)");
+
+// Case C: Existing contributions for node B do trigger suppression — verified implicitly by same filtering mechanism (matching contributions → effectiveContributionsForTarget.length > 0 → no deconstruct)
+assert(helperSrc.includes("effectiveContributionsForTarget.length") && helperSrc.includes("=== 0"), "Case C: matching contributions suppress redundant deconstruction (same guard, same check)");
+
+// Case D: Historical contributions remain present in returned state — postEpisodeContributions includes all effectiveContributions
+assert(helperSrc.includes("postEpisodeContributions: effectiveContributions") || helperSrc.includes("postEpisodeContributions : effectiveContributions"), "Case D: accumulated effectiveContributions (historical + new) returned via postEpisodeContributions");
+
 // ---- CU SYNTHESIS OBSERVABILITY (Batch 3 Iteration 1) ─────────────────────
 console.log("\n=== CU SYNTHESIS OBSERVABILITY ===\n");
 

tests/graph/live-update-experiment-helper.cjs

@@ -316,9 +316,18 @@ async function composeDeepInvestigationCheckpoint({ situationGraph, previousQues
 
   /* ── Canonical deconstruction: use focusedDeconstruct for Contribution derivation ─── */
   const effectiveContributions = contributions.length > 0 ? [...contributions] : [];
+
+  /* ── Target-scoped filtering: only contributions belonging to the current target suppress fresh deconstruction ─── */
+  let effectiveContributionsForTarget = effectiveContributions;
+  if (targetNodeId && effectiveContributions.length > 0) {
+    effectiveContributionsForTarget = effectiveContributions.filter(
+      (c) => c?.targetNodeId === targetNodeId,
+    );
+  }
+
   let newFindingsFromContributions = [];
 
-  if (effectiveContributions.length === 0 && answer && targetNodeId) {
+  if (effectiveContributionsForTarget.length === 0 && answer && targetNodeId) {
     const contribSeq = effectiveContributions.length + 1;
 
     /* ── Load production deconstruction seam (dynamic import to resolve ESM path aliases) ─── */