diff --git a/src/renderer/src/App.tsx b/src/renderer/src/App.tsx
index d1857e9d..fbed311b 100644
--- a/src/renderer/src/App.tsx
+++ b/src/renderer/src/App.tsx
@@ -295,28 +295,17 @@ function App(): React.JSX.Element {
     return () => window.removeEventListener('beforeunload', captureAndFlush)
   }, [])
 
-  // Periodically capture terminal scrollback buffers and persist to disk.
-  // Why: the shutdown path captures buffers in beforeunload, but periodic
-  // saves provide a safety net so scrollback is available on restart even
-  // if an unexpected exit (crash, force-kill) bypasses normal shutdown.
-  useEffect(() => {
-    const PERIODIC_SAVE_INTERVAL_MS = 3 * 60_000
-    const timer = window.setInterval(() => {
-      if (!useAppStore.getState().workspaceSessionReady || shutdownBufferCaptures.size === 0) {
-        return
-      }
-      for (const capture of shutdownBufferCaptures) {
-        try {
-          capture()
-        } catch {
-          // Don't let one pane's failure block the rest.
-        }
-      }
-      const state = useAppStore.getState()
-      void window.api.session.set(buildWorkspaceSessionPayload(state))
-    }, PERIODIC_SAVE_INTERVAL_MS)
-    return () => window.clearInterval(timer)
-  }, [])
+  // Why there is no periodic scrollback save: PR #461 added a 3-minute
+  // setInterval that re-serialized every mounted TerminalPane's scrollback
+  // so a crash wouldn't lose in-session output. With many panes of
+  // accumulated output, each tick blocked the renderer main thread for
+  // several seconds (serialize is synchronous and does a binary search on
+  // >512KB buffers), causing visible input lag across the whole app.
+  // The durable replacement is the out-of-process terminal daemon
+  // (PR #729), which preserves buffers across renderer crashes with no
+  // main-thread work. Non-daemon users lose in-session scrollback on an
+  // unexpected exit — an acceptable tradeoff vs. periodic UI stalls, and
+  // in line with how most terminal apps behave.
 
   useEffect(() => {
     if (!persistedUIReady) {
diff --git a/tests/e2e/helpers/orca-restart.ts b/tests/e2e/helpers/orca-restart.ts
new file mode 100644
index 00000000..6a7dc3ff
--- /dev/null
+++ b/tests/e2e/helpers/orca-restart.ts
@@ -0,0 +1,213 @@
+/**
+ * Two-launch Electron helper for restart-persistence tests.
+ *
+ * Why: the default `orcaPage` fixture creates a fresh `userDataDir` per test
+ * and deletes it on close, which is incompatible with a test that needs to
+ * quit the app and relaunch against the *same* on-disk state. This helper
+ * owns the shared userDataDir and gives each caller an `app`+`page` pair
+ * wired to it.
+ */
+
+import {
+  _electron as electron,
+  expect,
+  type ElectronApplication,
+  type Page,
+  type TestInfo
+} from '@stablyai/playwright-test'
+import { execSync } from 'child_process'
+import { existsSync, mkdtempSync, rmSync } from 'fs'
+import os from 'os'
+import path from 'path'
+
+type LaunchedOrca = {
+  app: ElectronApplication
+  page: Page
+}
+
+type RestartSession = {
+  userDataDir: string
+  launch: () => Promise<LaunchedOrca>
+  /** Gracefully close a launch, letting beforeunload flush session state. */
+  close: (app: ElectronApplication) => Promise<void>
+  /** Remove the shared userDataDir after the test is done. */
+  dispose: () => void
+}
+
+function shouldLaunchHeadful(testInfo: TestInfo): boolean {
+  return testInfo.project.metadata.orcaHeadful === true
+}
+
+function launchEnv(userDataDir: string, headful: boolean): NodeJS.ProcessEnv {
+  const { ELECTRON_RUN_AS_NODE: _unused, ...cleanEnv } = process.env
+  void _unused
+  return {
+    ...cleanEnv,
+    NODE_ENV: 'development',
+    ORCA_E2E_USER_DATA_DIR: userDataDir,
+    ...(headful ? { ORCA_E2E_HEADFUL: '1' } : { ORCA_E2E_HEADLESS: '1' })
+  }
+}
+
+/**
+ * Create a restart session tied to a persistent userDataDir.
+ *
+ * Why: keep the launch wiring identical to the shared fixture (mainPath,
+ * env stripping, headful toggle) so behavior differences between fixtures
+ * don't leak in as false positives for persistence bugs.
+ */
+export function createRestartSession(testInfo: TestInfo): RestartSession {
+  const mainPath = path.join(process.cwd(), 'out', 'main', 'index.js')
+  const userDataDir = mkdtempSync(path.join(os.tmpdir(), 'orca-e2e-restart-'))
+  const headful = shouldLaunchHeadful(testInfo)
+
+  const launch = async (): Promise<LaunchedOrca> => {
+    const app = await electron.launch({
+      args: [mainPath],
+      env: launchEnv(userDataDir, headful)
+    })
+    const page = await app.firstWindow({ timeout: 120_000 })
+    await page.waitForLoadState('domcontentloaded')
+    await page.waitForFunction(() => Boolean(window.__store), null, { timeout: 30_000 })
+    return { app, page }
+  }
+
+  const close = async (app: ElectronApplication): Promise<void> => {
+    // Why: mirror the shared fixture's shutdown race — give Electron 10s to run
+    // before-quit/will-quit (which drives the beforeunload → session.setSync
+    // flush this suite relies on) and only then fall back to SIGKILL.
+    const proc = app.process()
+    try {
+      await Promise.race([
+        app.close(),
+        new Promise<never>((_, reject) => {
+          setTimeout(() => reject(new Error('Timed out closing Electron app')), 10_000)
+        })
+      ])
+    } catch {
+      if (proc) {
+        try {
+          proc.kill('SIGKILL')
+        } catch {
+          /* already dead */
+        }
+      }
+    }
+  }
+
+  const dispose = (): void => {
+    if (existsSync(userDataDir)) {
+      rmSync(userDataDir, { recursive: true, force: true })
+    }
+  }
+
+  return { userDataDir, launch, close, dispose }
+}
+
+/**
+ * Attach a repo to the running renderer and wait until a terminal tab is
+ * active on its worktree. Matches the shared fixture's setup path so the
+ * first-launch state lines up with what real users persist.
+ */
+export async function attachRepoAndOpenTerminal(page: Page, repoPath: string): Promise<string> {
+  if (!isValidGitRepo(repoPath)) {
+    throw new Error(`attachRepoAndOpenTerminal: ${repoPath} is not a git repo`)
+  }
+
+  await page.evaluate(async (repoPath) => {
+    await window.api.repos.add({ path: repoPath })
+  }, repoPath)
+
+  await page.evaluate(async () => {
+    const store = window.__store
+    if (!store) {
+      return
+    }
+    await store.getState().fetchRepos()
+  })
+
+  await page.evaluate(async () => {
+    const store = window.__store
+    if (!store) {
+      return
+    }
+    const repos = store.getState().repos
+    for (const repo of repos) {
+      await store.getState().fetchWorktrees(repo.id)
+    }
+  })
+
+  await page.waitForFunction(
+    () => window.__store?.getState().workspaceSessionReady === true,
+    null,
+    { timeout: 30_000 }
+  )
+
+  // Why: fetchWorktrees() is async. Awaiting the outer page.evaluate returns
+  // before the Zustand worktree slice has observed the hydrated state, so a
+  // single evaluate() that reads worktreesByRepo can see an empty map. Poll
+  // the store until *any* worktree shows up, then pick the one that matches
+  // our seeded repo. Matching by basename is sufficient because each test
+  // run uses a unique tmp-dir suffix, and it avoids symlink-canonicalization
+  // mismatches between the path we pass in and the one the store records.
+  await expect
+    .poll(
+      async () =>
+        page.evaluate(() => {
+          const store = window.__store
+          if (!store) {
+            return false
+          }
+          return Object.values(store.getState().worktreesByRepo).flat().length > 0
+        }),
+      {
+        timeout: 15_000,
+        message: 'attachRepoAndOpenTerminal: seeded worktree never surfaced in the store'
+      }
+    )
+    .toBe(true)
+
+  const repoBasename = repoPath.split('/').filter(Boolean).pop() ?? ''
+  const worktreeId = await page.evaluate((repoBasename: string) => {
+    const store = window.__store
+    if (!store) {
+      return null
+    }
+    const state = store.getState()
+    const allWorktrees = Object.values(state.worktreesByRepo).flat()
+    // Why: the primary worktree lives at the repo root, so its path ends with
+    // the repo directory name. The secondary worktree is a sibling directory
+    // and will not match this suffix. This gives us the primary deterministically
+    // without depending on boolean fields on the worktree record.
+    const primary =
+      allWorktrees.find((worktree) => worktree.path.endsWith(`/${repoBasename}`)) ?? allWorktrees[0]
+    if (!primary) {
+      return null
+    }
+    state.setActiveWorktree(primary.id)
+    return primary.id
+  }, repoBasename)
+
+  if (!worktreeId) {
+    throw new Error('attachRepoAndOpenTerminal: test repo did not surface in the store')
+  }
+
+  return worktreeId
+}
+
+function isValidGitRepo(repoPath: string): boolean {
+  if (!repoPath || !existsSync(repoPath)) {
+    return false
+  }
+  try {
+    return (
+      execSync('git rev-parse --is-inside-work-tree', {
+        cwd: repoPath,
+        stdio: 'pipe',
+        encoding: 'utf8'
+      }).trim() === 'true'
+    )
+  } catch {
+    return false
+  }
+}
diff --git a/tests/e2e/terminal-restart-persistence.spec.ts b/tests/e2e/terminal-restart-persistence.spec.ts
new file mode 100644
index 00000000..87da9ab2
--- /dev/null
+++ b/tests/e2e/terminal-restart-persistence.spec.ts
@@ -0,0 +1,277 @@
+/**
+ * E2E tests for terminal scrollback persistence across clean app restarts.
+ *
+ * Why this suite exists:
+ *   PR #461 added a 3-minute periodic interval that re-serialized every
+ *   mounted TerminalPane's scrollback so an unclean exit (crash, SIGKILL)
+ *   wouldn't lose in-session output. With many panes of accumulated output,
+ *   each tick blocked the renderer main thread for seconds, causing visible
+ *   input lag across the whole app. The periodic save was removed in favor
+ *   of the out-of-process terminal daemon (PR #729). For users who don't
+ *   opt into the daemon, the `beforeunload` save in App.tsx is now the
+ *   *only* thing that preserves scrollback across a restart — this suite
+ *   locks that behavior down so a future regression can't silently return
+ *   us to "quit → empty terminal on relaunch."
+ *
+ * What it covers:
+ *   - Scrollback survives clean quit → relaunch (primary regression test).
+ *   - Tab layout (active worktree, terminal tab count) survives restart.
+ *   - Idle session writes stay infrequent (catches a reintroduced frequent
+ *     interval before it ships; weaker than asserting the 3-minute cadence
+ *     is gone, but doesn't require a minutes-long test run).
+ *
+ * What it does NOT try to cover:
+ *   - Main-thread input-lag improvement — machine-dependent and flaky.
+ *   - Crash/SIGKILL recovery — non-daemon users now intentionally lose
+ *     in-session scrollback on unclean exit; that's the tradeoff the
+ *     removed periodic save represented.
+ */
+
+import { readFileSync, existsSync } from 'fs'
+import type { ElectronApplication, Page } from '@stablyai/playwright-test'
+import { test, expect } from './helpers/orca-app'
+import { TEST_REPO_PATH_FILE } from './global-setup'
+import {
+  discoverActivePtyId,
+  execInTerminal,
+  waitForActiveTerminalManager,
+  waitForTerminalOutput,
+  waitForPaneCount,
+  getTerminalContent
+} from './helpers/terminal'
+import {
+  waitForSessionReady,
+  waitForActiveWorktree,
+  getActiveWorktreeId,
+  getWorktreeTabs,
+  ensureTerminalVisible
+} from './helpers/store'
+import { attachRepoAndOpenTerminal, createRestartSession } from './helpers/orca-restart'
+
+// Why: each test in this file does a full quit→relaunch cycle, which spawns
+// two Electron instances back-to-back. Running in serial keeps the isolated
+// userDataDirs from competing for the same Electron cache lock on cold start
+// and keeps the failure mode interpretable when something goes wrong.
+test.describe.configure({ mode: 'serial' })
+
+/**
+ * Shared bootstrap for a *first* launch: attach the seeded test repo,
+ * activate its worktree, ensure a terminal is mounted, and return the
+ * PTY id we can drive with `execInTerminal`.
+ *
+ * Why: every test in this file needs the exact same starting state on the
+ * first launch. Inlining it would obscure the thing each test is actually
+ * asserting about the *second* launch.
+ */
+async function bootstrapFirstLaunch(
+  page: Page,
+  repoPath: string
+): Promise<{ worktreeId: string; ptyId: string }> {
+  const worktreeId = await attachRepoAndOpenTerminal(page, repoPath)
+  await waitForSessionReady(page)
+  await waitForActiveWorktree(page)
+  await ensureTerminalVisible(page)
+
+  const hasPaneManager = await waitForActiveTerminalManager(page, 30_000)
+    .then(() => true)
+    .catch(() => false)
+  test.skip(
+    !hasPaneManager,
+    'Electron automation in this environment never mounts the TerminalPane manager, so restart-persistence assertions would only fail on harness setup.'
+  )
+  await waitForPaneCount(page, 1, 30_000)
+
+  const ptyId = await discoverActivePtyId(page)
+  return { worktreeId, ptyId }
+}
+
+/**
+ * Shared bootstrap for a *second* launch: just wait for the session to
+ * restore, and confirm the previously-active worktree is the active one
+ * again so downstream assertions operate against the right worktree.
+ */
+async function bootstrapRestoredLaunch(page: Page, expectedWorktreeId: string): Promise<void> {
+  await waitForSessionReady(page)
+  await expect
+    .poll(async () => getActiveWorktreeId(page), { timeout: 10_000 })
+    .toBe(expectedWorktreeId)
+  await ensureTerminalVisible(page)
+  // Why: the PaneManager remounts asynchronously after session hydration. The
+  // restored terminal surface is what we're about to assert against, so make
+  // sure it exists before any content/layout assertion races.
+  await waitForActiveTerminalManager(page, 30_000)
+  await waitForPaneCount(page, 1, 30_000)
+}
+
+test.describe('Terminal restart persistence', () => {
+  test('scrollback survives clean quit and relaunch', async (// oxlint-disable-next-line no-empty-pattern -- Playwright's second fixture arg is testInfo; the first must be an object destructure to opt out of the default fixture set.
+  {}, testInfo) => {
+    const repoPath = readFileSync(TEST_REPO_PATH_FILE, 'utf-8').trim()
+    if (!repoPath || !existsSync(repoPath)) {
+      test.skip(true, 'Global setup did not produce a seeded test repo')
+      return
+    }
+
+    const session = createRestartSession(testInfo)
+    let firstApp: ElectronApplication | null = null
+    let secondApp: ElectronApplication | null = null
+
+    try {
+      // ── First launch ────────────────────────────────────────────────
+      const firstLaunch = await session.launch()
+      firstApp = firstLaunch.app
+      const { worktreeId, ptyId } = await bootstrapFirstLaunch(firstLaunch.page, repoPath)
+
+      // Why: the marker must be distinctive enough that it can't appear in the
+      // restored prompt banner or a stray OSC sequence. The timestamp suffix
+      // keeps it unique across retries, and the trailing newline ensures the
+      // buffer snapshot contains it on a line of its own.
+      const marker = `SCROLLBACK_PERSIST_${Date.now()}`
+      await execInTerminal(firstLaunch.page, ptyId, `echo ${marker}`)
+      await waitForTerminalOutput(firstLaunch.page, marker)
+
+      // Why: closing the app triggers beforeunload → session.setSync, which is
+      // the one remaining codepath that flushes serialized scrollback to disk
+      // for non-daemon users. This is the behavior the suite is guarding.
+      await session.close(firstApp)
+      firstApp = null
+
+      // ── Second launch ───────────────────────────────────────────────
+      const secondLaunch = await session.launch()
+      secondApp = secondLaunch.app
+      await bootstrapRestoredLaunch(secondLaunch.page, worktreeId)
+
+      // Why: buffer restore replays the serialized output through xterm.write
+      // during pane mount. Poll the live terminal content rather than hitting
+      // the store because the store only sees the raw saved buffer, whereas
+      // restoreScrollbackBuffers can strip alt-screen sequences before write.
+      await expect
+        .poll(async () => (await getTerminalContent(secondLaunch.page)).includes(marker), {
+          timeout: 15_000,
+          message: 'Restored terminal did not contain the pre-quit scrollback marker'
+        })
+        .toBe(true)
+    } finally {
+      if (secondApp) {
+        await session.close(secondApp)
+      }
+      if (firstApp) {
+        await session.close(firstApp)
+      }
+      session.dispose()
+    }
+  })
+
+  test('active worktree and terminal tab count survive restart', async (// oxlint-disable-next-line no-empty-pattern -- Playwright's second fixture arg is testInfo; the first must be an object destructure to opt out of the default fixture set.
+  {}, testInfo) => {
+    const repoPath = readFileSync(TEST_REPO_PATH_FILE, 'utf-8').trim()
+    if (!repoPath || !existsSync(repoPath)) {
+      test.skip(true, 'Global setup did not produce a seeded test repo')
+      return
+    }
+
+    const session = createRestartSession(testInfo)
+    let firstApp: ElectronApplication | null = null
+    let secondApp: ElectronApplication | null = null
+
+    try {
+      // ── First launch ────────────────────────────────────────────────
+      const firstLaunch = await session.launch()
+      firstApp = firstLaunch.app
+      const { worktreeId } = await bootstrapFirstLaunch(firstLaunch.page, repoPath)
+
+      // Add a second terminal tab so the restart has layout state to restore
+      // beyond "one default tab." createTab goes through the same store path
+      // as the Cmd+T shortcut but doesn't depend on window focus timing.
+      await firstLaunch.page.evaluate((worktreeId: string) => {
+        const store = window.__store
+        if (!store) {
+          return
+        }
+        store.getState().createTab(worktreeId)
+      }, worktreeId)
+
+      await expect
+        .poll(async () => (await getWorktreeTabs(firstLaunch.page, worktreeId)).length, {
+          timeout: 5_000
+        })
+        .toBeGreaterThanOrEqual(2)
+
+      const tabsBefore = await getWorktreeTabs(firstLaunch.page, worktreeId)
+
+      await session.close(firstApp)
+      firstApp = null
+
+      // ── Second launch ───────────────────────────────────────────────
+      const secondLaunch = await session.launch()
+      secondApp = secondLaunch.app
+      await bootstrapRestoredLaunch(secondLaunch.page, worktreeId)
+
+      // Why: checking tab *count* (not ids) is the stable assertion — tab ids
+      // are regenerated on each launch because the renderer mints them fresh,
+      // while the persisted layout only carries the tab positions. Count
+      // survives; id identity does not.
+      await expect
+        .poll(async () => (await getWorktreeTabs(secondLaunch.page, worktreeId)).length, {
+          timeout: 10_000
+        })
+        .toBe(tabsBefore.length)
+    } finally {
+      if (secondApp) {
+        await session.close(secondApp)
+      }
+      if (firstApp) {
+        await session.close(firstApp)
+      }
+      session.dispose()
+    }
+  })
+
+  test('idle session does not spam session.set writes', async (// oxlint-disable-next-line no-empty-pattern -- Playwright's second fixture arg is testInfo; the first must be an object destructure to opt out of the default fixture set.
+  {}, testInfo) => {
+    const repoPath = readFileSync(TEST_REPO_PATH_FILE, 'utf-8').trim()
+    if (!repoPath || !existsSync(repoPath)) {
+      test.skip(true, 'Global setup did not produce a seeded test repo')
+      return
+    }
+
+    const session = createRestartSession(testInfo)
+    let app: ElectronApplication | null = null
+
+    try {
+      const { app: launchedApp, page } = await session.launch()
+      app = launchedApp
+      await bootstrapFirstLaunch(page, repoPath)
+
+      // Why: the periodic scrollback save that this branch removes was a
+      // `session.set` call on every tick. Counting `session.set` calls over a
+      // short idle window is a cheap proxy for "no high-frequency background
+      // writer was reintroduced." The 10s window intentionally stays below the
+      // per-test budget; the threshold is deliberately loose so normal user-
+      // driven store activity (tab auto-create, worktree activation) doesn't
+      // flake the test, while still catching an interval that fires every
+      // couple of seconds.
+      const callCount = await page.evaluate(async () => {
+        const api = (
+          window as unknown as { api: { session: { set: (...args: unknown[]) => unknown } } }
+        ).api
+        let count = 0
+        const originalSet = api.session.set.bind(api.session)
+        api.session.set = (...args: unknown[]) => {
+          count += 1
+          return originalSet(...args)
+        }
+        await new Promise((resolve) => setTimeout(resolve, 10_000))
+        api.session.set = originalSet
+        return count
+      })
+
+      expect(callCount).toBeLessThan(20)
+    } finally {
+      if (app) {
+        await session.close(app)
+      }
+      session.dispose()
+    }
+  })
+})