diff --git a/.changeset/fn-059-activity-log-failure-dedup.md b/.changeset/fn-059-activity-log-failure-dedup.md
new file mode 100644
index 0000000000..2e6af4e669
--- /dev/null
+++ b/.changeset/fn-059-activity-log-failure-dedup.md
@@ -0,0 +1,7 @@
+---
+"@runfusion/fusion": patch
+---
+
+summary: Prevent duplicate Task Failed entries after subsequent task updates.
+category: fix
+dev: Records failure activity only on a non-failed-to-failed task transition.
diff --git a/packages/core/src/__tests__/postgres/activity-log-parity.pg.test.ts b/packages/core/src/__tests__/postgres/activity-log-parity.pg.test.ts
index f656b71806..d44bc83d44 100644
--- a/packages/core/src/__tests__/postgres/activity-log-parity.pg.test.ts
+++ b/packages/core/src/__tests__/postgres/activity-log-parity.pg.test.ts
@@ -63,6 +63,56 @@ pgDescribe("activity log parity (PostgreSQL)", () => {
     });
   });
 
+  it("records one failure activity row per failed episode", async () => {
+    const store = h.store();
+    const task = await store.createTask({
+      title: "Deduplicate failed activity",
+      description: "Failure activity must only record episode transitions",
+    });
+    const error = "BLOCKED: prerequisite unavailable";
+
+    await store.updateTask(task.id, { status: "failed", error });
+    await store.updateTask(task.id, { title: "Deduplicate failed activity (updated)" });
+    await store.updateTask(task.id, { priority: "high" });
+
+    await vi.waitFor(async () => {
+      const failures = await store.getActivityLog({ type: "task:failed" });
+      expect(failures.filter((entry) => entry.taskId === task.id && entry.metadata?.error === error)).toHaveLength(1);
+    });
+
+    // An identical activity row in another project must not influence this project's feed.
+    await h.adminDb().insert(schema.project.activityLog).values({
+      projectId: "other-project",
+      id: "other-project-failure",
+      timestamp: new Date().toISOString(),
+      type: "task:failed",
+      taskId: task.id,
+      details: `Task ${task.id} failed: ${error}`,
+      metadata: { error },
+    });
+    expect((await store.getActivityLog({ type: "task:failed" }))
+      .filter((entry) => entry.taskId === task.id && entry.metadata?.error === error)).toHaveLength(1);
+
+    await store.updateTask(task.id, { status: null });
+    await store.updateTask(task.id, { status: "failed", error });
+    await store.updateTask(task.id, { description: "Still in the second failure episode" });
+
+    await vi.waitFor(async () => {
+      const failures = await store.getActivityLog({ type: "task:failed" });
+      expect(failures.filter((entry) => entry.taskId === task.id && entry.metadata?.error === error)).toHaveLength(2);
+    });
+
+    const errorless = await store.createTask({ description: "Errorless failed activity" });
+    await store.updateTask(errorless.id, { status: "failed" });
+    await store.updateTask(errorless.id, { title: "Errorless failed activity (updated)" });
+    await vi.waitFor(async () => {
+      const failures = await store.getActivityLog({ type: "task:failed" });
+      expect(failures.filter((entry) => entry.taskId === errorless.id)).toEqual([
+        expect.objectContaining({ metadata: undefined }),
+      ]);
+    });
+  });
+
   it("keeps failed writes best-effort so audit storage cannot break product operations", async () => {
     const layer = h.layer();
     const insert = vi.spyOn(layer.db, "insert").mockImplementation(() => {
diff --git a/packages/core/src/store.ts b/packages/core/src/store.ts
index e79ac7f37e..29df47da64 100644
--- a/packages/core/src/store.ts
+++ b/packages/core/src/store.ts
@@ -205,7 +205,7 @@ export interface TaskStoreEvents {
   second argument. The first argument remains the Task so existing one-argument subscribers are
   unchanged; absent metadata is unknown, never a legacy-lane claim.
   */
-  "task:updated": [task: Task, meta?: { lanes?: TaskMoveLanes }];
+  "task:updated": [task: Task, meta?: { lanes?: TaskMoveLanes; failedTransition?: boolean }];
   /*
   FNXC:CrossProcessDeleteObservation 2026-08-01-11:39:
   Observed outbox delivery is at-least-once, including a crash-window duplicate. The explicit
@@ -608,12 +608,19 @@ export class TaskStore extends EventEmitter<TaskStoreEvents> {
   FNXC:WorkflowEvents 2026-08-01-06:28:
   Safe lifecycle emission invokes listeners directly to isolate listener failures, so it bypasses
   EventEmitter.emit. Decorate this path too; otherwise the hot update surfaces silently miss lanes.
+
+  FNXC:ActivityLogFailureDedup 2026-08-19-20:03:
+  Update metadata now carries the failure-transition marker, but warm-cache updates must retain the
+  existing resolved lanes decoration and cold-cache updates must preserve absent metadata.
   */
   public emitTaskLifecycleEventSafely( event: "task:created" | "task:updated" | "task:deleted", args: TaskStoreEvents["task:created"] | TaskStoreEvents["task:updated"] | TaskStoreEvents["task:deleted"], ): boolean {
-    if (event === "task:updated" && args.length === 1) {
+    if (event === "task:updated") {
       const task = args[0] as Task;
+      const metadata = args[1] as TaskStoreEvents["task:updated"][1] | undefined;
       const lanes = this.laneCache.get(task.id);
-      if (lanes !== undefined) return emitTaskLifecycleEventSafelyImpl(this, event, [task, { lanes }]);
+      if (lanes !== undefined && metadata?.lanes === undefined) {
+        return emitTaskLifecycleEventSafelyImpl(this, event, [task, { ...metadata, lanes }]);
+      }
     }
     return emitTaskLifecycleEventSafelyImpl(this, event, args);
   }
diff --git a/packages/core/src/task-store/lifecycle-ops.ts b/packages/core/src/task-store/lifecycle-ops.ts
index be504967e1..499af8381e 100644
--- a/packages/core/src/task-store/lifecycle-ops.ts
+++ b/packages/core/src/task-store/lifecycle-ops.ts
@@ -318,9 +318,14 @@ export function setupActivityLogListenersImpl(store: TaskStore): void {
       );
     });
 
-    // Task updated (check for failures)
-    store.on("task:updated", (task) => {
-      if (task.status === "failed") {
+    /*
+    FNXC:ActivityLogFailureDedup 2026-08-19-19:33:
+    A failed-state bookkeeping update is not a new durable failure episode. Only the serialized
+    non-failed → failed transition from updateTask may append task:failed activity, so listeners
+    cannot turn subsequent task mutations into an unbounded project.activity_log write path.
+    */
+    store.on("task:updated", (task, meta) => {
+      if (meta?.failedTransition) {
         store.recordActivityFromListener(
           {
             type: "task:failed",
diff --git a/packages/core/src/task-store/task-update.ts b/packages/core/src/task-store/task-update.ts
index bd68029fbe..32e48db046 100644
--- a/packages/core/src/task-store/task-update.ts
+++ b/packages/core/src/task-store/task-update.ts
@@ -1245,7 +1245,11 @@ export async function updateTaskUnlockedImpl(store: TaskStore, id: string, updat
           lanes: respecifyMoveLanes,
         });
       }
-      store.emitTaskLifecycleEventSafely("task:updated", [task]);
+      const failedTransition = !wasFailed && task.status === "failed";
+      store.emitTaskLifecycleEventSafely("task:updated", [
+        task,
+        failedTransition ? { failedTransition: true } : undefined,
+      ]);
       return task;
     }
   }