diff --git a/apps/backend/src/services/worker-platform/worker-hall-service.ts b/apps/backend/src/services/worker-platform/worker-hall-service.ts
index a78b65cb..cddaa352 100644
--- a/apps/backend/src/services/worker-platform/worker-hall-service.ts
+++ b/apps/backend/src/services/worker-platform/worker-hall-service.ts
@@ -26,6 +26,15 @@ import { WorkerListCache } from './worker-list-cache.js'
 const HALL_LIST_CACHE_TTL_MS = 2_000
 const hallListCache = new WorkerListCache(HALL_LIST_CACHE_TTL_MS)
 
+// 排行榜每次都要对全量 work_orders + work_order_shares 做去重聚合（生产实测
+// 742ms / 35,483 块 ≈ 277MB），而结果只在工单验收完成时才变化。
+// 大厅每次刷新/窗口聚焦都会触发前端重新拉取，不缓存时该查询独占全库约 8% 的缓冲读。
+// TTL 取 30s：对排行榜而言足够新鲜（名次变化以分钟计），又能把重算频率压掉两个数量级。
+// 注意：结果里包含"当前打手自己的名次"（SQL 里 OR worker_id = $2），缓存键必须带上
+// workerId，否则会把 A 的名次错发给 B。
+const HALL_LEADERBOARD_CACHE_TTL_MS = 30_000
+const hallLeaderboardCache = new WorkerListCache(HALL_LEADERBOARD_CACHE_TTL_MS)
+
 export async function listWorkerHallOrders(query: JsonObject = {}, session: WorkerSession) {
   requireActiveWorkerSession(session)
   const page = normalizePage(query.page)
@@ -117,11 +126,14 @@ export async function getWorkerHallCategories(session: WorkerSession) {
 export async function getWorkerHallLeaderboard(session: WorkerSession, period: unknown = 'all') {
   requireActiveWorkerSession(session)
   const leaderboardPeriod = period === 'daily' ? 'daily' : 'all'
-  const entries = await listWorkerCompletionLeaderboard({
-    limit: 10,
-    currentWorkerId: session.workerId,
-    period: leaderboardPeriod,
-  })
+  const cacheKey = JSON.stringify([leaderboardPeriod, Number(session.workerId)])
+  const entries = await hallLeaderboardCache.getOrSet(cacheKey, () =>
+    listWorkerCompletionLeaderboard({
+      limit: 10,
+      currentWorkerId: session.workerId,
+      period: leaderboardPeriod,
+    }),
+  )
   const current = entries.find((entry) => entry.workerId === session.workerId)
   return {
     items: entries
diff --git a/apps/backend/src/services/worker-platform/worker-platform-service.test.ts b/apps/backend/src/services/worker-platform/worker-platform-service.test.ts
index 01015891..2ebb2476 100644
--- a/apps/backend/src/services/worker-platform/worker-platform-service.test.ts
+++ b/apps/backend/src/services/worker-platform/worker-platform-service.test.ts
@@ -2691,7 +2691,7 @@ test('打手端「我的订单」概览必须有覆盖索引 + 短缓存（数
   // 短缓存：只缓存读取，会话校验与逾期结算必须每次照常执行
   const service = await readFile(new URL('./worker-my-order-service.ts', import.meta.url), 'utf8')
   assert.ok(
-    service.includes('myOrderOverviewCache.getOrSet(String(worker.id)'),
+    service.includes('myOrderOverviewCache.getOrSet('),
     '概览缺少短缓存：调用量 0.5 次/秒，每次都要重算',
   )
   const overviewBody = service.slice(
@@ -2707,6 +2707,49 @@ test('打手端「我的订单」概览必须有覆盖索引 + 短缓存（数
   )
 })
 
+test('「我的订单」缓存必须事件驱动失效（长 TTL 的安全前提）', async () => {
+  const { readFile } = await import('node:fs/promises')
+
+  // 缓存实例独立成中立模块：realtime 要引用失效入口，而 my-order-service 已 import realtime，
+  // 实例放任何一方都会形成循环依赖。
+  const cacheModule = await readFile(new URL('./worker-my-order-cache.ts', import.meta.url), 'utf8')
+  assert.ok(cacheModule.includes('invalidateWorkerMyOrderCaches'), '缺少按打手批量失效的入口')
+  assert.ok(
+    cacheModule.includes('evictByPrefix'),
+    '失效必须能按键前缀整体逐出，否则只能等 TTL 自然过期',
+  )
+  // 键必须带统一前缀 ["w",workerId,...]，前缀失效才不会误伤其它打手
+  assert.ok(
+    cacheModule.includes(`['w', Number(workerId) || 0, scope, detail]`),
+    '缓存键必须以打手 id 为前缀，事件失效才能按打手整体逐出',
+  )
+
+  // 失效必须挂在实时快照广播里：那里恰好已有工单 + 全部拼单人，受影响集合最完整，
+  // 且 21 处工单写入路径（含取消申请/反馈处理）都会经过。
+  const realtime = await readFile(
+    new URL('../realtime/realtime-event-service.ts', import.meta.url),
+    'utf8',
+  )
+  assert.ok(
+    realtime.includes('invalidateWorkerMyOrderCaches(['),
+    '实时快照广播必须触发「我的订单」缓存失效',
+  )
+  const snapshotBody = realtime.slice(
+    realtime.indexOf('async function publishWorkOrderSnapshots'),
+    realtime.indexOf('async function publishWorkerFinanceSnapshots'),
+  )
+  for (const field of [
+    'workOrder.assigned_worker_id',
+    'workOrder.last_assigned_worker_id',
+    'shares.map((share) => share.worker_id)',
+  ]) {
+    assert.ok(
+      snapshotBody.includes(field),
+      `受影响打手集合缺少 ${field}：漏掉会导致该打手最长 30s 看到旧数据`,
+    )
+  }
+})
+
 test('按运营态筛选的计数必须有覆盖索引（生产实测每次读 248MB）', async () => {
   const { readFile } = await import('node:fs/promises')
   const migration = await readFile(
@@ -2895,3 +2938,30 @@ test('平台概览必须有短缓存（一次并发 5 条 count，池只有 8 
     '概览必须走 getOrSet 合并并发请求',
   )
 })
+
+test('接单排行榜必须有短缓存（单次全表聚合 277MB/742ms，独占全库约 8% 缓冲读）', async () => {
+  const { readFile } = await import('node:fs/promises')
+  const src = await readFile(new URL('./worker-hall-service.ts', import.meta.url), 'utf8')
+
+  assert.ok(
+    src.includes('new WorkerListCache(HALL_LEADERBOARD_CACHE_TTL_MS)'),
+    '排行榜缺少短缓存：大厅每次刷新/聚焦都会重新对全量工单做去重聚合',
+  )
+  assert.ok(
+    src.includes('hallLeaderboardCache.getOrSet('),
+    '排行榜必须走 getOrSet（同时合并并发请求）',
+  )
+
+  // 结果包含"当前打手自己的名次"（SQL 里 OR worker_id = $2），
+  // 缓存键漏了 workerId 会把 A 的名次错发给 B —— 这是正确性问题，不是性能问题。
+  const leaderboardBody = src.slice(
+    src.indexOf('export async function getWorkerHallLeaderboard'),
+    src.indexOf('export function maskWorkerLeaderboardDisplayName'),
+  )
+  assert.ok(
+    leaderboardBody.includes('session.workerId') &&
+      /cacheKey/.test(leaderboardBody) &&
+      leaderboardBody.includes('JSON.stringify'),
+    '排行榜缓存键必须包含 workerId 与 period',
+  )
+})