diff --git a/dbms/src/Common/TiFlashMetrics.h b/dbms/src/Common/TiFlashMetrics.h
index 7d29442957d..7496107d889 100644
--- a/dbms/src/Common/TiFlashMetrics.h
+++ b/dbms/src/Common/TiFlashMetrics.h
@@ -543,6 +543,7 @@ static_assert(RAFT_REGION_BIG_WRITE_THRES * 4 < RAFT_REGION_BIG_WRITE_MAX, "Inva
       "Raft learner read failure reason counter",                                                                                   \
       Counter,                                                                                                                      \
       F(type_request_error, {{"type", "request_error"}}),                                                                           \
+      F(type_request_error_legacy, {{"type", "request_error_legacy"}}),                                                             \
       F(type_read_index_timeout, {{"type", "read_index_timeout"}}),                                                                 \
       F(type_not_found_tiflash, {{"type", "not_found_tiflash"}}),                                                                   \
       F(type_epoch_not_match, {{"type", "epoch_not_match"}}),                                                                       \
diff --git a/dbms/src/Storages/KVStore/Read/Proxy.cpp b/dbms/src/Storages/KVStore/Read/Proxy.cpp
index d8aaf2e3d39..6a416cb8953 100644
--- a/dbms/src/Storages/KVStore/Read/Proxy.cpp
+++ b/dbms/src/Storages/KVStore/Read/Proxy.cpp
@@ -23,6 +23,8 @@ BatchReadIndexRes TiFlashRaftProxyHelper::batchReadIndex(
     return batchReadIndex_v2(req, timeout_ms);
 }
 
+// Learner Read will use `ReadIndexDataNode` which will send read index requests in batch,
+// so this function is not actually used after the optimization.
 BatchReadIndexRes TiFlashRaftProxyHelper::batchReadIndex_v2(
     const std::vector<kvrpcpb::ReadIndexRequest> & req,
     uint64_t timeout_ms) const
@@ -38,10 +40,10 @@ BatchReadIndexRes TiFlashRaftProxyHelper::batchReadIndex_v2(
     {
         if (auto task = makeReadIndexTask(r); !task)
         {
-            // The read index request is not sent successfully.
-            GET_METRIC(tiflash_raft_learner_read_failures_count, type_request_error).Increment();
             kvrpcpb::ReadIndexResponse res;
             res.mutable_region_error();
+            // The read index request is not sent successfully.
+            GET_METRIC(tiflash_raft_learner_read_failures_count, type_request_error_legacy).Increment();
             resps.emplace_back(std::move(res), r.context().region_id());
         }
         else
diff --git a/dbms/src/Storages/KVStore/Read/ReadIndexDataNode.cpp b/dbms/src/Storages/KVStore/Read/ReadIndexDataNode.cpp
index 2e36c041b40..ddc942a0403 100644
--- a/dbms/src/Storages/KVStore/Read/ReadIndexDataNode.cpp
+++ b/dbms/src/Storages/KVStore/Read/ReadIndexDataNode.cpp
@@ -102,6 +102,7 @@ void ReadIndexDataNode::runOneRound(const TiFlashRaftProxyHelper & helper, const
                 else
                 {
                     TEST_LOG_FMT("failed to make ReadIndexTask for region_id={} ts {}", region_id, max_ts);
+                    GET_METRIC(tiflash_raft_learner_read_failures_count, type_request_error).Increment();
                     run_it = running_tasks.try_emplace(max_ts, region_id, max_ts).first;
                     run_it->second.resp.mutable_region_error();
                 }
@@ -182,8 +183,8 @@ void ReadIndexDataNode::ReadIndexElement::doPoll(
                 TEST_LOG_FMT("poll ReadIndexElement timeout for region_id={}", region_id);
 
                 clean_task = true;
-                resp.mutable_region_error()
-                    ->mutable_server_is_busy(); // set region_error `server_is_busy` for task timeout
+                // set region_error `server_is_busy` for task timeout
+                resp.mutable_region_error()->mutable_server_is_busy();
             }
             else
             {
diff --git a/dbms/src/Storages/KVStore/tests/gtest_async_tasks.cpp b/dbms/src/Storages/KVStore/tests/gtest_async_tasks.cpp
index 1f57eed56cb..1160137161f 100644
--- a/dbms/src/Storages/KVStore/tests/gtest_async_tasks.cpp
+++ b/dbms/src/Storages/KVStore/tests/gtest_async_tasks.cpp
@@ -28,16 +28,18 @@ TEST(AsyncTasksTest, AsyncTasksNormal)
     auto log = DB::Logger::get();
     LOG_INFO(log, "Cancel and addTask");
     // Cancel and addTask
+    // 3 -> 1 -> 4 -> 2
     {
         auto async_tasks = std::make_unique<TestAsyncTasks>(1, 1, 2);
         auto m = std::make_shared<std::mutex>();
-        auto m2 = std::make_shared<std::mutex>();
         int flag = 0;
         std::unique_lock cl(*m);
-        async_tasks->addTask(1, [m, &flag, &async_tasks, &m2]() {
+        std::atomic_bool finished_flag = false;
+        async_tasks->addTask(1, [m, &flag, &async_tasks, &finished_flag]() {
             auto cancel_handle = async_tasks->getCancelHandleFromExecutor(1);
-            std::scoped_lock rl2(*m2);
-            std::scoped_lock rl(*m);
+            std::scoped_lock rl(*m); // 2
+            SCOPE_EXIT({ finished_flag.store(true); });
+            // Run after `cl` is released.
             if (cancel_handle->isCanceled())
             {
                 return;
@@ -47,8 +49,15 @@ TEST(AsyncTasksTest, AsyncTasksNormal)
         async_tasks->asyncCancelTask(1);
         ASSERT_FALSE(async_tasks->isScheduled(1));
         async_tasks->addTask(1, [&flag]() { flag = 2; });
-        cl.unlock();
-        std::scoped_lock rl2(*m2);
+        cl.unlock(); // Now can task 1 run.
+        int count = 0;
+        using namespace std::chrono_literals;
+        while (!finished_flag.load())
+        {
+            count += 1;
+            ASSERT(count < 6);
+            std::this_thread::sleep_for(200ms);
+        }
         ASSERT_NO_THROW(async_tasks->fetchResult(1));
         ASSERT_EQ(flag, 2);
     }