diff --git a/.agents/notes/implemented/feature/2026-07-16-persistent-pty-sessions.i18n.yaml b/.agents/notes/implemented/feature/2026-07-16-persistent-pty-sessions.i18n.yaml
index f772b6d5e6..c4fa994098 100644
--- a/.agents/notes/implemented/feature/2026-07-16-persistent-pty-sessions.i18n.yaml
+++ b/.agents/notes/implemented/feature/2026-07-16-persistent-pty-sessions.i18n.yaml
@@ -2,5 +2,5 @@
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
 #   pnpm run verify-translation-pairing --write
-2026-07-16-persistent-pty-sessions.md: 8a9ec669a064924ecdb9693fff5c4cafea59d90c
-2026-07-16-persistent-pty-sessions.zh.md: 157bee408d041b5c373dd2a4d41f82de90a67e0e
+2026-07-16-persistent-pty-sessions.md: fdab27d73258cdcc7382e52fad09f749d26ae890
+2026-07-16-persistent-pty-sessions.zh.md: a0cac7b835c2b5c20d6569e87d49b7fdc5b4dac2
diff --git a/.agents/notes/implemented/feature/2026-07-16-persistent-pty-sessions.md b/.agents/notes/implemented/feature/2026-07-16-persistent-pty-sessions.md
index 8a9ec669a0..fdab27d732 100644
--- a/.agents/notes/implemented/feature/2026-07-16-persistent-pty-sessions.md
+++ b/.agents/notes/implemented/feature/2026-07-16-persistent-pty-sessions.md
@@ -92,7 +92,7 @@ Background sends use the existing task completion notice and `task_output` resul
 
 The top-level `node-pty` child is the ownership anchor. On close, the backend stops callbacks, snapshots its transitive descendants by parent PID in children-first order, sends `SIGTERM`, waits, rescans for children forked during shutdown, sends `SIGKILL` to the remaining descendant tree, and verifies that every non-zombie descendant left the process table while the shell is still alive. A matching Linux zombie has no executable work and therefore counts as quiescent, allowing shell shutdown to reap or reparent it. Only then does the backend stop the shell with its own TERM/grace/KILL sequence. Every captured PID includes process-start identity so reuse cannot redirect escalation.
 
-Teardown reports root exit and survivor cleanup independently. It does not claim success merely because the shell exited; disposal resolves only after no captured non-quiescent tree member remains or returns a cleanup failure naming the survivors. A failed close is not cached forever: the registry and local session retain their closing fence but allow a later close to retry after the external survivor condition changes. Service disposal still clears its backend, reservation, and owner-detacher registries when a close fails. It never broadens ownership to every member of the root PID's POSIX session.
+Teardown reports root exit and survivor cleanup independently. It does not claim success merely because the shell exited; disposal resolves only after no captured non-quiescent tree member remains or returns a cleanup failure naming the survivors. A failed close is not cached forever: the registry and local session each clear the fence only when it still names that failed attempt, so a later explicit or lifecycle close retries after the external survivor condition changes without disturbing a newer concurrent attempt. Service disposal still clears its backend, reservation, and owner-detacher registries when a close fails. It never broadens ownership to every member of the root PID's POSIX session.
 
 ### Composition and rollout
 
diff --git a/.agents/notes/implemented/feature/2026-07-16-persistent-pty-sessions.zh.md b/.agents/notes/implemented/feature/2026-07-16-persistent-pty-sessions.zh.md
index 157bee408d..a0cac7b835 100644
--- a/.agents/notes/implemented/feature/2026-07-16-persistent-pty-sessions.zh.md
+++ b/.agents/notes/implemented/feature/2026-07-16-persistent-pty-sessions.zh.md
@@ -92,7 +92,7 @@ Tier 2 在持续 `idleSilenceMs` 没有输出后返回 `inferred_idle`，因此
 
 顶层 `node-pty` 子进程是所有权锚点。关闭时，后端先停止 callback，再按父 PID 以子进程优先顺序捕获其传递子进程、发送 `SIGTERM` 并等待，然后重新扫描关停期间 fork 出的子进程，向剩余子孙进程树发送 `SIGKILL`，并在 shell 仍存活时验证每个非僵尸子孙进程都已离开进程表。身份匹配的 Linux 僵尸进程已无可执行工作，因此视为静止；shell 关闭时会回收它或将其重新挂接给负责回收的父进程。完成这些步骤后，后端才用 shell 自身的 TERM、宽限等待、KILL 序列停止 shell。每个捕获的 PID 都包含进程启动身份，避免 PID 复用把升级信号发给无关进程。
 
-teardown 独立报告根进程退出与存活进程清理。它不会只因 shell 退出就声称成功；dispose 只有在已捕获的进程树中不再存在非静止成员后才完成，否则返回清理失败并列出存活者。失败的 close 不会永久缓存：注册表与本地会话会保留关闭围栏，但在外部存活进程状态改变后允许后续 close 重试。即使某个 close 失败，服务 dispose 仍会清空其后端、预留与 owner detacher 注册表。所有权绝不会扩大到根 PID 所属 POSIX 会话的全部成员。
+teardown 独立报告根进程退出与存活进程清理。它不会只因 shell 退出就声称成功；dispose 只有在已捕获的进程树中不再存在非静止成员后才完成，否则返回清理失败并列出存活者。失败的 close 不会永久缓存：注册表与本地会话各自仅在关闭围栏仍指向该次失败尝试时才将其清除，因此外部存活进程状态改变后，后续的显式 close 或生命周期 close 会重试，且不会干扰较新的并发尝试。即使某个 close 失败，服务 dispose 仍会清空其后端、预留与 owner detacher 注册表。所有权绝不会扩大到根 PID 所属 POSIX 会话的全部成员。
 
 ### 组合与推行
 
diff --git a/packages/pty/pty/README.md b/packages/pty/pty/README.md
index 4213a3458f..798991d5fd 100644
--- a/packages/pty/pty/README.md
+++ b/packages/pty/pty/README.md
@@ -13,7 +13,7 @@ Owner-scoped persistent PTY seam. `PtyService` registers as `ctx.pty`, mints opa
 - A successful spawn publishes one `PtySessionId`. The optional `name` is owner-local display metadata, never authority.
 - One session accepts at most one live send operation. Reads and signals may observe it; another send fails until the operation settles.
 - `PtySendResult.waitReason` and `sessionStatus` are independent. `session_exit` describes the top-level PTY process, not an arbitrary foreground command.
-- `kill()` and disposal resolve only after the backend's captured process tree is quiescent. A cleanup failure rejects instead of claiming success and leaves the close retriable.
+- `kill()` and disposal resolve only after the backend's captured process tree is quiescent. A cleanup failure rejects instead of claiming success and clears the matching backend and registry fences so a later close can retry without disturbing a newer attempt.
 
 The seam contains no `node-pty`, sandbox, tool-schema, prompt, task, or terminal-rendering policy. Implementations own terminal mechanics; consumers own model presentation and optional background-task registration.
 
diff --git a/packages/pty/pty/src/index.ts b/packages/pty/pty/src/index.ts
index 47c2a2662d..c7896d5113 100644
--- a/packages/pty/pty/src/index.ts
+++ b/packages/pty/pty/src/index.ts
@@ -457,8 +457,14 @@ export class PtyService extends Service {
     const results = await Promise.allSettled(records.map(async (record) => {
       const closing = record.closing ?? record.session.close(reason)
       record.closing = closing
-      await closing
-      this.sessions.delete(record.id)
+      try {
+        await closing
+        this.sessions.delete(record.id)
+      } catch (error: unknown) {
+        // A concurrent retry may already own a newer fence; never clear it.
+        if (record.closing === closing) record.closing = undefined
+        throw error
+      }
     }))
     const failures = results
       .filter((result): result is PromiseRejectedResult => result.status === 'rejected')
diff --git a/packages/pty/pty/tests/service.spec.ts b/packages/pty/pty/tests/service.spec.ts
index caa255ac5b..bca131011d 100644
--- a/packages/pty/pty/tests/service.spec.ts
+++ b/packages/pty/pty/tests/service.spec.ts
@@ -540,8 +540,15 @@ describe('PtyService ownership and lifecycle', () => {
       sessions: Map<PtySessionIdType, unknown>
       closeRecords(records: unknown[], reason: string): Promise<void>
     }
-    await expect(internal.closeRecords([...internal.sessions.values()], 'test failure')).rejects.toThrow('failed to close 1 PTY session')
+    const records = [...internal.sessions.values()]
+    const firstFailure = expect(internal.closeRecords(records, 'test failure')).rejects.toThrow('failed to close 1 PTY session')
+    const joinedFailure = expect(internal.closeRecords(records, 'joined failure')).rejects.toThrow('failed to close 1 PTY session')
+    await firstFailure
+    await joinedFailure
     b.sessions[0]!.rejectClose = false
+    await expect(internal.closeRecords([...internal.sessions.values()], 'retry')).resolves.toBeUndefined()
+    expect(b.sessions[0]!.closed).toEqual(['test failure', 'retry'])
+    expect(internal.sessions.size).toBe(0)
     await disposePtyService(ctx)
     await expect(service.spawn(owner, { type: 'stub' })).rejects.toMatchObject({ code: 'SERVICE_DISPOSING' })
   })