diff --git a/desktop/src-tauri/src/commands/host_move_tests.rs b/desktop/src-tauri/src/commands/host_move_tests.rs
index 93038c4641..42fe37dce9 100644
--- a/desktop/src-tauri/src/commands/host_move_tests.rs
+++ b/desktop/src-tauri/src/commands/host_move_tests.rs
@@ -238,3 +238,34 @@ fn destination_rejection_reports_source_stopped_never_moved_or_restarted() {
     assert_eq!(f.store.journal.sent.len(), 2);
     super::super::host_start::validate_journal(&f.store, &f.owner, RELAY).unwrap();
 }
+
+#[test]
+fn renewed_stop_receipt_preserves_observation_and_releases_move_once() {
+    let mut f = Fixture::new();
+    f.receipt(Outcome::Stopped);
+    let intent = f.store.journal.moves[&f.id].clone();
+    let stop = f.store.journal.sent.get_mut(&intent.stop).unwrap();
+    let command = stop.command.clone();
+    let original = stop.receipt.clone().unwrap();
+    super::super::host_start_store::prepare_receipt(stop, &f.owner, &f.source, RELAY, &[], 1200)
+        .unwrap();
+    let renewed = stop.receipt.as_ref().unwrap();
+    assert_ne!(renewed.id, original.id);
+    assert_eq!(renewed.content, original.content);
+    assert_eq!(renewed.tags, original.tags);
+    assert_eq!(stop.command, command);
+    let req = request(&f.owner, stop, RELAY).unwrap();
+    let observation =
+        host_execution::decrypt_receipt(&f.owner, &stop.registration, renewed, &command, &req)
+            .unwrap();
+    assert_eq!(observation.observed_at, 110);
+    assert_eq!(observation.outcome, Outcome::Stopped);
+    f.store.save().unwrap();
+    drop(f.store);
+    f.store = Store::open_dir(f.dir.path(), &f.owner.public_key().to_hex(), RELAY).unwrap();
+    release_to_outbox(&mut f.store, &f.owner, RELAY, &f.id, &intent, 1201).unwrap();
+    let released = f.store.journal.moves[&f.id].clone();
+    assert!(release_to_outbox(&mut f.store, &f.owner, RELAY, &f.id, &released, 1202).is_err());
+    assert_eq!(f.store.journal.sent.len(), 2);
+    super::super::host_start::validate_journal(&f.store, &f.owner, RELAY).unwrap();
+}
diff --git a/desktop/src-tauri/src/host_move_tracer.rs b/desktop/src-tauri/src/host_move_tracer.rs
index fbada8200a..017b1bd8b7 100644
--- a/desktop/src-tauri/src/host_move_tracer.rs
+++ b/desktop/src-tauri/src/host_move_tracer.rs
@@ -347,7 +347,7 @@ pub(super) async fn trace(
                     "ordinary-stop-success.json",
                     &serde_json::to_value(peer_stop).map_err(|_| "Stop status")?,
                 )?;
-                exercise_ordinary_restart(app, root, &peer, relay_url, &peer_run)?;
+                exercise_ordinary_restart(app, root, &peer, relay_url, &peer_run).await?;
                 save(root, "move-finish", &json!({"done":true}))?;
                 cleanup(app)?;
                 return Ok(());
@@ -361,7 +361,7 @@ pub(super) async fn trace(
 
 // Exercise the existing pair Start/Restart entry points, not a fixture-only
 // ledger shortcut. A stale Stop retry must not label a live successor stopped.
-fn exercise_ordinary_restart(
+async fn exercise_ordinary_restart(
     app: &tauri::AppHandle,
     root: &Path,
     peer: &str,
@@ -387,6 +387,12 @@ fn exercise_ordinary_restart(
     {
         return Err("old successful Stop retry misreported successor as stopped".into());
     }
+    save(
+        root,
+        "ordinary-start-success.json",
+        &serde_json::to_value(&started).map_err(|_| "Start status")?,
+    )?;
+    wait_live_run(app, peer, relay_url, &started_run).await?;
     let restarted = managed_agents::restart_managed_agent_runtime(
         peer.into(),
         relay_url.into(),
@@ -400,6 +406,7 @@ fn exercise_ordinary_restart(
     if restarted_run == started_run || restarted.pid.is_none() {
         return Err("ordinary Restart did not create a fresh runtime".into());
     }
+    wait_live_run(app, peer, relay_url, &restarted_run).await?;
     let stopped = managed_agents::stop_managed_agent_runtime(
         peer.into(),
         relay_url.into(),
@@ -416,6 +423,36 @@ fn exercise_ordinary_restart(
     )
 }
 
+// Spawn is explicitly not Ready. Exercise the happy path only after observing
+// this exact new runtime on the relay; an immediate startup Stop may fail closed.
+async fn wait_live_run(
+    app: &tauri::AppHandle,
+    agent: &str,
+    relay_url: &str,
+    run: &str,
+) -> Result<(), String> {
+    let state = app.state::<AppState>();
+    let owner = state.signing_keys()?.public_key().to_hex();
+    for _ in 0..60 {
+        let runs = commands::get_presence_runs(
+            state.clone(),
+            owner.clone(),
+            relay_url.into(),
+            vec![agent.into()],
+        )
+        .await?;
+        let now = nostr::Timestamp::now().as_secs();
+        if runs.get(agent).is_some_and(|runs| {
+            runs.iter()
+                .any(|r| r.run == run && r.status != "offline" && r.expires_at > now)
+        }) {
+            return Ok(());
+        }
+        tokio::time::sleep(Duration::from_secs(1)).await;
+    }
+    Err("ordinary restart generation not observed live".into())
+}
+
 async fn pulse(
     app: &tauri::AppHandle,
     owner: &str,
diff --git a/desktop/src-tauri/src/host_start_tracer.rs b/desktop/src-tauri/src/host_start_tracer.rs
index ee69da67d4..df63d9d93d 100644
--- a/desktop/src-tauri/src/host_start_tracer.rs
+++ b/desktop/src-tauri/src/host_start_tracer.rs
@@ -91,6 +91,10 @@ pub fn run() -> Result<(), String> {
         .map_err(|_| "native app build failed")?;
     let handle = app.handle().clone();
     let relay = args[3].clone();
+    // Use run_return: Wry's non-returning run can discard AppHandle::exit's
+    // requested code. Success also requires that the trace actually completed.
+    let completed = std::sync::Arc::new(std::sync::atomic::AtomicBool::new(false));
+    let task_completed = completed.clone();
     tauri::async_runtime::spawn(async move {
         let result = if role.starts_with("move-") {
             move_trace::trace(&handle, &role, &root, &relay).await
@@ -98,15 +102,22 @@ pub fn run() -> Result<(), String> {
             trace(&handle, &role, &root, &relay).await
         };
         match result {
-            Ok(()) => handle.exit(0),
+            Ok(()) => {
+                task_completed.store(true, std::sync::atomic::Ordering::Release);
+                handle.exit(0);
+            }
             Err(error) => {
                 eprintln!("Start tracer failed: {error}");
                 handle.exit(1);
             }
         }
     });
-    app.run(|_, _| {});
-    Ok(())
+    let exit_code = app.run_return(|_, _| {});
+    if exit_code == 0 && completed.load(std::sync::atomic::Ordering::Acquire) {
+        Ok(())
+    } else {
+        Err("tracer did not complete successfully; inspect result files and logs".into())
+    }
 }
 
 fn keys(root: &Path, name: &str) -> Result<Keys, String> {
diff --git a/desktop/src/shared/api/tauriManagedAgents.ts b/desktop/src/shared/api/tauriManagedAgents.ts
index 838d908df0..12d1ced43f 100644
--- a/desktop/src/shared/api/tauriManagedAgents.ts
+++ b/desktop/src/shared/api/tauriManagedAgents.ts
@@ -117,8 +117,13 @@ export async function stopManagedAgentRuntime(
 export async function restartManagedAgentRuntime(
   pubkey: string,
   relayUrl: string,
+  selectedRunId?: string | null,
 ): Promise<ManagedAgentRuntimeStatus> {
-  return invokeTauri("restart_managed_agent_runtime", { pubkey, relayUrl });
+  return invokeTauri("restart_managed_agent_runtime", {
+    pubkey,
+    relayUrl,
+    selectedRunId,
+  });
 }
 
 export async function putManagedAgentRuntimeLifecycle(
diff --git a/docs/host-execution.md b/docs/host-execution.md
index 2fb34014ae..b83ccb9eb7 100644
--- a/docs/host-execution.md
+++ b/docs/host-execution.md
@@ -124,7 +124,10 @@ cannot be reported as stopped by replaying that earlier success.
 
 An explicit ordinary Start after confirmed exact Stop, and the existing pair
 Restart action, enter the same ledger for a fresh generation. Automatic reconcile
-remains fenced. Provider `!shutdown` behavior is unchanged. Config-edit, delete
+remains fenced. Spawn is not Ready: an immediate Stop during ACP pool initialization
+can return RootExited, even when partial children were drained. It cannot certify
+replacement; the actual tracer preserves this negative case rather than forcing
+Stopped. Provider `!shutdown` behavior is unchanged. Config-edit, delete
 and application cleanup retain their separate best-effort behavior and are not
 certified Stop claims. No new run authority or provider control channel is added.
 
@@ -260,8 +263,13 @@ Cleanup reaped only the fixture's tracked root, not a certified Stop.
 The 2026-08-31 Move tracer additionally completed confirmed selected-source Stop
 followed by destination spawn with the same agent and matching fresh run/host.
 The unrelated source peer survived Move and then completed its own ordinary Stop.
-An old source token was rejected at the destination. Both executors exited 0;
-fixture cleanup was not used as certification evidence.
+An old source token was rejected at the destination. The expanded tracer also
+exercised ordinary Start after confirmed Stop, waited for that exact new run's
+signed live presence, then used ordinary Restart and finally its own confirmed
+Stop. Retrying the earlier successful Stop could not report the successor stopped.
+Both executors completed with the required result files; fixture cleanup was not
+used as certification evidence. A debug tracer failure is checked from its result
+and log as well as exit code (the tracer uses the returning native event loop).
 
 Remaining gates include real-provider and second-physical-host validation,
 actual Hosts native UI capture, authenticated asynchronous lifecycle receipt