fix(apps): eliminate blank-app window on launch and add crash visibility

- start the apps server before createWindow instead of after the full
  service-init chain: the Apps view was reachable ~10s before port 3210
  was listening, so every app iframe opened to connection-refused
- retry EADDRINUSE binds (quick relaunch races left the server disabled
  for the whole session with no retry)
- app frame: auto-retry the iframe instead of a dead manual Retry, and
  surface when the apps server itself is down (new apps:serverStatus)
- log render-process-gone/child-process-gone reasons; renderer helper
  crashes were previously invisible in packaged builds

apps/x/apps/main/src/ipc.ts

@@ -1565,6 +1565,9 @@ export function setupIpcHandlers() {
       return qualifyAndDisconnectComposioGoogle();
     },
     // Rowboat Apps handlers (spec §13)
+    'apps:serverStatus': async () => {
+      return appsServer.getServerStatus();
+    },
     'apps:list': async () => {
       const status = appsServer.getServerStatus();
       const apps = await appsIndexer.listApps();

apps/x/apps/main/src/main.ts

@@ -342,6 +342,16 @@ function createWindow() {
   }
 }
 
+// Renderer/child process deaths are otherwise silent in packaged builds (the
+// window or an app iframe just goes blank). Log the reason so crash reports
+// can be correlated with what Chromium thought happened.
+app.on('render-process-gone', (_event, webContents, details) => {
+  console.error(`[Crash] renderer gone: reason=${details.reason} exitCode=${details.exitCode} url=${webContents.getURL()}`);
+});
+app.on('child-process-gone', (_event, details) => {
+  console.error(`[Crash] child process gone: type=${details.type} reason=${details.reason} exitCode=${details.exitCode ?? ''} name=${details.name ?? ''}`);
+});
+
 app.whenReady().then(async () => {
   // Register custom protocol before creating window.
   // In production this serves the renderer SPA; in dev (and prod) it also
@@ -383,6 +393,24 @@ app.whenReady().then(async () => {
   setupIpcHandlers();
   setupBrowserEventForwarding();
 
+  // Start the Rowboat Apps server (per-app origins on 127.0.0.1:3210) BEFORE
+  // the window and the long service-init chain below. The Apps view is
+  // reachable as soon as the window paints; starting the server last meant
+  // every app iframe hit connection-refused (blank app) for the first ~10s of
+  // each launch. Route registration and the token cipher are synchronous;
+  // the listen itself is fire-and-forget.
+  registerAppsHostApi();
+  // GitHub publish token at rest: encrypt via the OS keychain when available
+  // (core stays electron-free; the cipher is injected here).
+  setGithubTokenCipher({
+    isAvailable: () => safeStorage.isEncryptionAvailable(),
+    encrypt: (plain) => safeStorage.encryptString(plain).toString('base64'),
+    decrypt: (encrypted) => safeStorage.decryptString(Buffer.from(encrypted, 'base64')),
+  });
+  initAppsServer().catch((error) => {
+    console.error('[Apps] Failed to start:', error);
+  });
+
   createWindow();
 
   // Start workspace watcher as a main-process service
@@ -503,21 +531,6 @@ app.whenReady().then(async () => {
   // start chrome extension sync server
   initChromeSync();
 
-  // start the Rowboat Apps server (per-app origins on 127.0.0.1:3210) with the
-  // full Host API (tools/fetch/llm/copilot behind the capability gate)
-  registerAppsHostApi();
-
-  // GitHub publish token at rest: encrypt via the OS keychain when available
-  // (core stays electron-free; the cipher is injected here).
-  setGithubTokenCipher({
-    isAvailable: () => safeStorage.isEncryptionAvailable(),
-    encrypt: (plain) => safeStorage.encryptString(plain).toString('base64'),
-    decrypt: (encrypted) => safeStorage.decryptString(Buffer.from(encrypted, 'base64')),
-  });
-  initAppsServer().catch((error) => {
-    console.error('[Apps] Failed to start:', error);
-  });
-
   app.on("activate", () => {
     if (BrowserWindow.getAllWindows().length === 0) {
       createWindow();

apps/x/apps/renderer/src/components/apps/app-frame.tsx

@@ -1,5 +1,5 @@
 import { useEffect, useState } from 'react'
-import { ArrowLeft, BadgeCheck, ExternalLink, Info, RotateCw, UploadCloud } from 'lucide-react'
+import { ArrowLeft, BadgeCheck, ExternalLink, Info, Loader2, RotateCw, UploadCloud } from 'lucide-react'
 import type { rowboatApp } from '@x/shared'
 import { appOpened } from '@/lib/analytics'
 import { AppDetail } from '@/components/apps/app-detail'
@@ -36,6 +36,24 @@ export function AppFrame({ app, onBack }: { app: rowboatApp.AppSummary; onBack:
     return () => window.clearTimeout(timer)
   }, [watchKey])
 
+  // Stuck recovery: the usual cause is the apps server not being reachable yet
+  // (it starts with the main process; on a fresh launch or a quick relaunch the
+  // first iframe load can beat it). Auto-retry instead of demanding a manual
+  // click, and say when the server itself is the problem.
+  const [serverDown, setServerDown] = useState(false)
+  useEffect(() => {
+    if (loadState !== 'stuck') return
+    let cancelled = false
+    void window.ipc.invoke('apps:serverStatus', {}).then((s) => {
+      if (!cancelled) setServerDown(!s.running)
+    }).catch(() => { /* status probe is best-effort */ })
+    const timer = window.setTimeout(() => setReloadNonce((n) => n + 1), 4000)
+    return () => {
+      cancelled = true
+      window.clearTimeout(timer)
+    }
+  }, [loadState])
+
   return (
     <div className="flex h-full flex-col">
       <div className="flex items-center gap-2 border-b border-border px-3 py-2">
@@ -105,13 +123,18 @@ export function AppFrame({ app, onBack }: { app: rowboatApp.AppSummary; onBack:
           />
           {loadState === 'stuck' && (
             <div className="absolute inset-0 flex flex-col items-center justify-center gap-3 bg-background/95 text-sm">
-              <div className="text-muted-foreground">This app is taking too long to load.</div>
+              <div className="text-muted-foreground">
+                {serverDown ? 'The Rowboat apps server is still starting up.' : 'This app is taking too long to load.'}
+              </div>
+              <div className="flex items-center gap-2 text-xs text-muted-foreground">
+                <Loader2 className="size-3.5 animate-spin" /> Retrying automatically…
+              </div>
               <button
                 type="button"
                 onClick={() => setReloadNonce((n) => n + 1)}
                 className="rounded-md border border-border px-3 py-1.5 font-medium hover:bg-accent"
               >
-                Retry
+                Retry now
               </button>
               <div className="font-mono text-xs text-muted-foreground">{app.origin}</div>
             </div>

apps/x/packages/core/src/apps/server.ts

@@ -741,22 +741,29 @@ export async function init(): Promise<void> {
             startLagMonitor();
             await startWatcher();
             const expressApp = createApp();
-            await new Promise<void>((resolve, reject) => {
-                const s = expressApp.listen(APPS_PORT, '127.0.0.1', () => {
-                    server = s;
+            const listenOn = (host: string) => new Promise<Server>((resolve, reject) => {
+                const s = expressApp.listen(APPS_PORT, host, () => resolve(s));
+                s.on('error', (error: NodeJS.ErrnoException) => reject(error));
+            });
+            // EADDRINUSE almost always means a previous Rowboat instance is
+            // still shutting down and holding the port (quick relaunch). Retry
+            // on the SAME port for a while — never scan for alternate ports
+            // (§6.1), origins embed the port — instead of disabling apps for
+            // the whole session on the first failure.
+            for (let attempt = 1; ; attempt++) {
+                try {
+                    server = await listenOn('127.0.0.1');
                     serverError = null;
                     console.log(`[Apps] server on 127.0.0.1:${APPS_PORT} (host suffix ${APPS_HOST_SUFFIX}), dir ${APPS_DIR}`);
-                    resolve();
-                });
-                s.on('error', (error: NodeJS.ErrnoException) => {
-                    if (error.code === 'EADDRINUSE') {
-                        // Never scan for alternate ports (§6.1) — origins embed the port.
-                        reject(new Error(`Port ${APPS_PORT} is already in use.`));
-                        return;
-                    }
-                    reject(error);
-                });
-            });
+                    break;
+                } catch (error) {
+                    const code = (error as NodeJS.ErrnoException).code;
+                    if (code !== 'EADDRINUSE') throw error;
+                    if (attempt >= 15) throw new Error(`Port ${APPS_PORT} is already in use.`);
+                    if (attempt === 1) console.warn(`[Apps] port ${APPS_PORT} in use — retrying (old instance still shutting down?)`);
+                    await new Promise((r) => setTimeout(r, 1000));
+                }
+            }
             // Dual-stack loopback: also listen on ::1 (see server6 note above).
             // Best-effort — some machines have IPv6 disabled.
             await new Promise<void>((resolve) => {

apps/x/packages/shared/src/ipc.ts

@@ -1228,6 +1228,13 @@ const ipcSchemas = {
     }),
   },
   // Rowboat Apps (spec §13) — M1 local channels.
+  'apps:serverStatus': {
+    req: z.object({}),
+    res: z.object({
+      running: z.boolean(),
+      error: z.string().optional(),
+    }),
+  },
   'apps:list': {
     req: z.object({}),
     res: z.object({