fix: robust worker interaction

- Text heartbeat every 30s (not ping frames) to survive Cloudflare/proxy
- Comment on completed/failed workflow re-dispatches to worker with state resume
- Dispatch retries 3 times with 5s delay (survives reconnect window)

src/agent.rs

@@ -153,7 +153,16 @@ impl AgentManager {
                     require_plan_approval: false,
                 };
 
-                match self.worker_mgr.assign_workflow(assign).await {
+                // Retry dispatch up to 3 times (worker might be reconnecting)
+                let mut dispatch_result = self.worker_mgr.assign_workflow(assign.clone()).await;
+                for attempt in 1..3 {
+                    if dispatch_result.is_ok() { break; }
+                    tracing::warn!("Dispatch attempt {} failed, retrying in 5s...", attempt);
+                    tokio::time::sleep(std::time::Duration::from_secs(5)).await;
+                    dispatch_result = self.worker_mgr.assign_workflow(assign.clone()).await;
+                }
+
+                match dispatch_result {
                     Ok(name) => {
                         tracing::info!("Workflow {} dispatched to worker '{}'", workflow_id, name);
                     }
@@ -177,8 +186,71 @@ impl AgentManager {
                 }
             }
             AgentEvent::Comment { workflow_id, content } => {
-                if let Err(e) = self.worker_mgr.forward_comment(&workflow_id, &content).await {
-                    tracing::warn!("Failed to forward comment for workflow {}: {}", workflow_id, e);
+                // Try to forward to running worker first
+                if let Err(_) = self.worker_mgr.forward_comment(&workflow_id, &content).await {
+                    // No worker handling this workflow — re-dispatch it
+                    // Load workflow to get project_id and requirement
+                    let wf = sqlx::query_as::<_, crate::db::Workflow>(
+                        "SELECT * FROM workflows WHERE id = ?"
+                    ).bind(&workflow_id).fetch_optional(&self.pool).await.ok().flatten();
+
+                    if let Some(wf) = wf {
+                        tracing::info!("Re-dispatching workflow {} with comment", workflow_id);
+                        // Load latest state for resume
+                        let state_json: Option<String> = sqlx::query_scalar(
+                            "SELECT state_json FROM agent_state_snapshots WHERE workflow_id = ? ORDER BY created_at DESC LIMIT 1"
+                        ).bind(&workflow_id).fetch_optional(&self.pool).await.ok().flatten();
+
+                        let mut initial_state = state_json
+                            .and_then(|json| serde_json::from_str::<crate::state::AgentState>(&json).ok());
+
+                        // Attach comment as user feedback + reset failed/waiting steps
+                        if let Some(ref mut state) = initial_state {
+                            for step in &mut state.steps {
+                                if matches!(step.status, crate::state::StepStatus::Failed) {
+                                    step.status = crate::state::StepStatus::Pending;
+                                }
+                                if matches!(step.status, crate::state::StepStatus::WaitingUser) {
+                                    step.status = crate::state::StepStatus::Running;
+                                }
+                            }
+                            if let Some(order) = state.first_actionable_step() {
+                                if let Some(step) = state.steps.iter_mut().find(|s| s.order == order) {
+                                    step.user_feedbacks.push(content);
+                                }
+                            }
+                        }
+
+                        let assign = crate::worker::ServerToWorker::WorkflowAssign {
+                            workflow_id: workflow_id.clone(),
+                            project_id: wf.project_id.clone(),
+                            requirement: wf.requirement,
+                            template_id: if wf.template_id.is_empty() { None } else { Some(wf.template_id) },
+                            initial_state,
+                            require_plan_approval: false,
+                        };
+
+                        let btx = {
+                            let mut map = self.broadcast.write().await;
+                            map.entry(wf.project_id.clone())
+                                .or_insert_with(|| broadcast::channel(64).0)
+                                .clone()
+                        };
+
+                        match self.worker_mgr.assign_workflow(assign).await {
+                            Ok(name) => {
+                                let _ = sqlx::query("UPDATE workflows SET status = 'executing', status_reason = '' WHERE id = ?")
+                                    .bind(&workflow_id).execute(&self.pool).await;
+                                let _ = btx.send(WsMessage::WorkflowStatusUpdate {
+                                    workflow_id, status: "executing".into(),
+                                });
+                                tracing::info!("Workflow re-dispatched to worker '{}'", name);
+                            }
+                            Err(e) => {
+                                tracing::error!("Failed to re-dispatch workflow {}: {}", workflow_id, e);
+                            }
+                        }
+                    }
                 }
             }
         }

src/worker_runner.rs

@@ -92,19 +92,15 @@ async fn connect_and_run(
     // Store the new ws_tx so relay tasks can use it
     *shared_ws_tx.lock().await = Some(ws_tx);
 
-    // Ping keepalive
-    let ping_tx = shared_ws_tx.clone();
+    // Heartbeat keepalive (text message so Cloudflare/proxies don't drop it)
+    let hb_tx = shared_ws_tx.clone();
     let ping_task = tokio::spawn(async move {
         let mut interval = tokio::time::interval(std::time::Duration::from_secs(30));
         loop {
             interval.tick().await;
-            let guard = ping_tx.lock().await;
-            if guard.is_none() { break; }
-            // Can't send while holding mutex with Option, drop and re-acquire
-            drop(guard);
-            let mut guard = ping_tx.lock().await;
+            let mut guard = hb_tx.lock().await;
             if let Some(ref mut tx) = *guard {
-                if tx.send(Message::Ping(vec![].into())).await.is_err() {
+                if tx.send(Message::Text(r#"{"type":"heartbeat"}"#.into())).await.is_err() {
                     *guard = None;
                     break;
                 }

src/ws_worker.rs

@@ -90,6 +90,10 @@ async fn handle_worker_socket(socket: WebSocket, state: Arc<WsWorkerState>) {
         while let Some(Ok(msg)) = receiver.next().await {
             match msg {
                 Message::Text(text) => {
+                    // Fast-path: heartbeat
+                    if text.contains("\"heartbeat\"") {
+                        continue;
+                    }
                     match serde_json::from_str::<WorkerToServer>(&text) {
                         Ok(worker_msg) => {
                             handle_worker_message(&state_clone, worker_msg).await;