fix(desktop): retry failed initial relay dials (#2564)

Signed-off-by: npub12gtutshhh76rx0jx697f32f9tffd4hhp3hx58fp4x6u4uemkm7sqf8f757 <5217c5c2f7bfb4333e46d17c98a9255a52dadee18dcd43a43536b95e6776dfa0@buzz.block.builderlab.xyz>
Co-authored-by: npub12gtutshhh76rx0jx697f32f9tffd4hhp3hx58fp4x6u4uemkm7sqf8f757 <5217c5c2f7bfb4333e46d17c98a9255a52dadee18dcd43a43536b95e6776dfa0@buzz.block.builderlab.xyz>

desktop/src/shared/api/relayClientSession.ts

@@ -534,8 +534,6 @@ export class RelayClient {
   }
 
   private async connect() {
-    // Clear any pending stability timer from a previous connection — a new
-    // connect attempt resets the clock and must re-arm the timer on success.
     if (this.stabilityTimer !== null) {
       window.clearTimeout(this.stabilityTimer);
       this.stabilityTimer = null;
@@ -545,51 +543,66 @@ export class RelayClient {
       this.hasConnectedOnce ? "reconnecting" : "connecting",
     );
 
-    if (!this.relayUrl) {
-      this.relayUrl = await getRelayWsUrl();
-    }
-
     const generation = ++this.connectionGeneration;
     this.onMessageChannel = new Channel<unknown>((message) => {
-      void this.handleWsMessage(message, generation);
-    });
-
-    this.wsId = await invoke<number>("plugin:websocket|connect", {
-      url: this.relayUrl,
-      onMessage: this.onMessageChannel,
-      config: {},
-    });
-
-    await new Promise<void>((resolve, reject) => {
-      const timeout = window.setTimeout(() => {
-        this.authRequest = null;
+      void this.handleWsMessage(message, generation).catch((error) => {
+        if (generation !== this.connectionGeneration) return;
         this.resetConnection(
-          new Error("Timed out while waiting for relay authentication."),
+          this.normalizeRelayError(error, "Relay connection errored."),
         );
-        reject(new Error("Timed out while waiting for relay authentication."));
-      }, AUTH_TIMEOUT_MS);
-
-      this.authRequest = {
-        pendingEventId: "",
-        resolve,
-        reject,
-        timeout,
-      };
+      });
     });
 
-    // Start a stability timer instead of resetting backoff immediately.
-    // The backoff resets to its base value only after BACKOFF_RESET_STABLE_MS
-    // of uninterrupted uptime, preventing fast reconnect loops from erasing
-    // the exponential backoff that throttles them.
-    this.stabilityTimer = window.setTimeout(() => {
-      this.stabilityTimer = null;
-      this.reconnectDelayMs = RECONNECT_BASE_DELAY_MS;
-    }, BACKOFF_RESET_STABLE_MS);
+    try {
+      if (!this.relayUrl) {
+        this.relayUrl = await getRelayWsUrl();
+      }
+      const wsId = await invoke<number>("plugin:websocket|connect", {
+        url: this.relayUrl,
+        onMessage: this.onMessageChannel,
+        config: {},
+      });
+      if (generation !== this.connectionGeneration) {
+        void closeWebSocket(wsId, "stale connection attempt");
+        throw new Error("Relay connection attempt was superseded.");
+      }
+      this.wsId = wsId;
 
-    await this.replayLiveSubscriptions();
-    this.connectionStateEmitter.set("connected");
-    this.stallWatchdog.start();
-    this.emitReconnectIfNeeded();
+      await new Promise<void>((resolve, reject) => {
+        const timeout = window.setTimeout(() => {
+          const error = new Error("Relay authentication timed out.");
+          this.authRequest = null;
+          this.resetConnection(error);
+          reject(error);
+        }, AUTH_TIMEOUT_MS);
+
+        this.authRequest = {
+          pendingEventId: "",
+          resolve,
+          reject,
+          timeout,
+        };
+      });
+
+      this.stabilityTimer = window.setTimeout(() => {
+        this.stabilityTimer = null;
+        this.reconnectDelayMs = RECONNECT_BASE_DELAY_MS;
+      }, BACKOFF_RESET_STABLE_MS);
+
+      await this.replayLiveSubscriptions();
+      this.connectionStateEmitter.set("connected");
+      this.stallWatchdog.start();
+      this.emitReconnectIfNeeded();
+    } catch (error) {
+      const connectionError = this.normalizeRelayError(
+        error,
+        "Failed to connect to relay.",
+      );
+      if (generation === this.connectionGeneration) {
+        this.resetConnection(connectionError);
+      }
+      throw connectionError;
+    }
   }
 
   private async subscribe(

desktop/src/testing/e2eBridge.ts

@@ -250,6 +250,8 @@ type E2eConfig = {
     openerError?: string;
     /** Delay binding signatures so specs can exercise request supersession. */
     nostrBindSignDelayMs?: number;
+    /** Reject successive mock WebSocket connect attempts, then resume. */
+    websocketConnectErrors?: string[];
     stallWebsocketSends?: boolean;
     userSearchDelayMs?: number;
     // NIP-IA gate inputs — see tests/helpers/bridge.ts:MockBridgeOptions for
@@ -8498,6 +8500,11 @@ async function connectRealSocket(args: { url?: string; onMessage: unknown }) {
 }
 
 async function connectMockSocket(args: { onMessage: unknown }) {
+  const connectError = getConfig()?.mock?.websocketConnectErrors?.shift();
+  if (connectError) {
+    throw new Error(connectError);
+  }
+
   if (mockWebsocketSendMutexWedged) {
     return new Promise<number>(() => {});
   }

desktop/tests/e2e/relay-reconnect.spec.ts

@@ -80,6 +80,33 @@ test.beforeEach(async ({ page }) => {
   await installMockBridge(page);
 });
 
+test("failed initial relay dial retries automatically", async ({ page }) => {
+  await installMockBridge(page, {
+    websocketConnectErrors: ["mock relay pod unavailable"],
+  });
+  await page.goto("/");
+
+  // App-shell preconnect owns a keep-alive request. The first native dial is
+  // rejected before a socket ID exists; the session must still enter its
+  // backoff loop and recover without a click, query, or reload.
+  await expect
+    .poll(
+      () =>
+        page.evaluate(() => {
+          const getState = (
+            window as Window & {
+              __BUZZ_E2E_GET_RELAY_CONNECTION_STATE__?: () => string;
+            }
+          ).__BUZZ_E2E_GET_RELAY_CONNECTION_STATE__;
+          if (!getState) throw new Error("Relay state seam is not installed.");
+          return getState();
+        }),
+      { timeout: 10_000 },
+    )
+    .toBe("connected");
+  await expect(page.getByTestId("channel-general")).toBeVisible();
+});
+
 test("passive relay watchdog does not write while the websocket is half-open", async ({
   page,
 }) => {

desktop/tests/helpers/bridge.ts

@@ -261,6 +261,8 @@ type MockBridgeOptions = {
   openerError?: string;
   /** Delay binding signatures so specs can exercise request supersession. */
   nostrBindSignDelayMs?: number;
+  /** Reject successive mock WebSocket connect attempts, then resume. */
+  websocketConnectErrors?: string[];
   stallWebsocketSends?: boolean;
   userSearchDelayMs?: number;
   /**