diff --git a/cdc/kv/client.go b/cdc/kv/client.go
index 0ba2f18c28a..9c169817cc3 100644
--- a/cdc/kv/client.go
+++ b/cdc/kv/client.go
@@ -69,6 +69,8 @@ const (
 	// failed region will be reloaded via `BatchLoadRegionsWithKeyRange` API. So we
 	// don't need to force reload region anymore.
 	regionScheduleReload = false
+
+	scanRegionsConcurrency = 1024
 )
 
 // time interval to force kv client to terminate gRPC stream and reconnect
@@ -424,6 +426,8 @@ func (s *eventFeedSession) eventFeed(ctx context.Context, ts uint64) error {
 	})
 
 	g.Go(func() error {
+		g, ctx := errgroup.WithContext(ctx)
+		g.SetLimit(scanRegionsConcurrency)
 		for {
 			select {
 			case <-ctx.Done():
@@ -997,8 +1001,7 @@ func (s *eventFeedSession) receiveFromStream(
 
 	// always create a new region worker, because `receiveFromStream` is ensured
 	// to call exactly once from outer code logic
-	worker := newRegionWorker(s.changefeed, s, addr)
-
+	worker := newRegionWorker(s.changefeed, s, addr, pendingRegions)
 	defer worker.evictAllRegions()
 
 	g.Go(func() error {
diff --git a/cdc/kv/region_worker.go b/cdc/kv/region_worker.go
index 0f1e5eb63b9..08fd2a3c956 100644
--- a/cdc/kv/region_worker.go
+++ b/cdc/kv/region_worker.go
@@ -112,10 +112,13 @@ type regionWorker struct {
 
 	// how many pending input events
 	inputPending int32
+
+	pendingRegions *syncRegionFeedStateMap
 }
 
 func newRegionWorker(
 	changefeedID model.ChangeFeedID, s *eventFeedSession, addr string,
+	pendingRegions *syncRegionFeedStateMap,
 ) *regionWorker {
 	metrics := &regionWorkerMetrics{}
 	metrics.metricReceivedEventSize = eventSize.WithLabelValues("received")
@@ -149,6 +152,8 @@ func newRegionWorker(
 		concurrency:   s.client.config.WorkerConcurrent,
 		metrics:       metrics,
 		inputPending:  0,
+
+		pendingRegions: pendingRegions,
 	}
 }
 
@@ -184,7 +189,7 @@ func (w *regionWorker) checkShouldExit() error {
 	empty := w.checkRegionStateEmpty()
 	// If there is no region maintained by this region worker, exit it and
 	// cancel the gRPC stream.
-	if empty {
+	if empty && w.pendingRegions.len() == 0 {
 		w.cancelStream(time.Duration(0))
 		return cerror.ErrRegionWorkerExit.GenWithStackByArgs()
 	}
diff --git a/cdc/kv/region_worker_test.go b/cdc/kv/region_worker_test.go
index fef317d4c0d..c8fcebe4a18 100644
--- a/cdc/kv/region_worker_test.go
+++ b/cdc/kv/region_worker_test.go
@@ -154,7 +154,7 @@ func TestRegionWokerHandleEventEntryEventOutOfOrder(t *testing.T) {
 		regionspan.ToComparableSpan(span),
 		0, &tikv.RPCContext{}), 0)
 	state.start()
-	worker := newRegionWorker(model.ChangeFeedID{}, s, "")
+	worker := newRegionWorker(model.ChangeFeedID{}, s, "", newSyncRegionFeedStateMap())
 	require.Equal(t, 2, cap(worker.outputCh))
 
 	// Receive prewrite2 with empty value.
@@ -313,7 +313,7 @@ func TestRegionWorkerHandleEventsBeforeStartTs(t *testing.T) {
 		rpcCtx:     &tikv.RPCContext{},
 	}, 0)
 	s1.start()
-	w := newRegionWorker(model.ChangeFeedID{}, s, "")
+	w := newRegionWorker(model.ChangeFeedID{}, s, "", newSyncRegionFeedStateMap())
 
 	err := w.handleResolvedTs(ctx, &resolvedTsEvent{
 		resolvedTs: 5,
diff --git a/cdc/processor/sinkmanager/tasks.go b/cdc/processor/sinkmanager/tasks.go
index c6b6ea8757b..a736b6fca50 100644
--- a/cdc/processor/sinkmanager/tasks.go
+++ b/cdc/processor/sinkmanager/tasks.go
@@ -33,8 +33,8 @@ var (
 	maxUpdateIntervalSize = defaultMaxUpdateIntervalSize
 
 	// Sink manager schedules table tasks based on lag. Limit the max task range
-	// can be helpful to reduce changefeed latency.
-	maxTaskRange = 5 * time.Second
+	// can be helpful to reduce changefeed latency for large initial data.
+	maxTaskRange = 30 * time.Minute
 )
 
 // Used to record the progress of the table.