diff --git a/hack/benchmark-go.sh b/hack/benchmark-go.sh
index 2090e47448..5512a8b60c 100755
--- a/hack/benchmark-go.sh
+++ b/hack/benchmark-go.sh
@@ -20,4 +20,4 @@ set -o pipefail
 
 KUBE_ROOT=$(dirname "${BASH_SOURCE}")/..
 
-KUBE_COVER="" KUBE_RACE=" " "${KUBE_ROOT}/hack/test-go.sh" -- -test.run="^X" -benchtime=1s -bench=. -benchmem
+KUBE_COVER="" KUBE_RACE=" " "${KUBE_ROOT}/hack/test-go.sh" -- -test.run="^X" -benchtime=1s -bench=. -benchmem $@
diff --git a/pkg/client/cache/reflector.go b/pkg/client/cache/reflector.go
index 32cca298da..340d73a135 100644
--- a/pkg/client/cache/reflector.go
+++ b/pkg/client/cache/reflector.go
@@ -106,17 +106,24 @@ var (
 	errorStopRequested = errors.New("Stop requested")
 )
 
-// resyncChan returns a channel which will receive something when a resync is required.
-func (r *Reflector) resyncChan() <-chan time.Time {
+// resyncChan returns a channel which will receive something when a resync is
+// required, and a cleanup function.
+func (r *Reflector) resyncChan() (<-chan time.Time, func() bool) {
 	if r.resyncPeriod == 0 {
-		return neverExitWatch
+		return neverExitWatch, func() bool { return false }
 	}
-	return time.After(r.resyncPeriod)
+	// The cleanup function is required: imagine the scenario where watches
+	// always fail so we end up listing frequently. Then, if we don't
+	// manually stop the timer, we could end up with many timers active
+	// concurrently.
+	t := time.NewTimer(r.resyncPeriod)
+	return t.C, t.Stop
 }
 
 func (r *Reflector) listAndWatch(stopCh <-chan struct{}) {
 	var resourceVersion string
-	resyncCh := r.resyncChan()
+	resyncCh, cleanup := r.resyncChan()
+	defer cleanup()
 
 	list, err := r.listerWatcher.List()
 	if err != nil {
diff --git a/pkg/client/cache/reflector_test.go b/pkg/client/cache/reflector_test.go
index 54bfc136d3..a63dfd0356 100644
--- a/pkg/client/cache/reflector_test.go
+++ b/pkg/client/cache/reflector_test.go
@@ -18,6 +18,7 @@ package cache
 
 import (
 	"fmt"
+	"math/rand"
 	"strconv"
 	"testing"
 	"time"
@@ -95,7 +96,8 @@ func TestRunUntil(t *testing.T) {
 func TestReflector_resyncChan(t *testing.T) {
 	s := NewStore(MetaNamespaceKeyFunc)
 	g := NewReflector(&testLW{}, &api.Pod{}, s, time.Millisecond)
-	a, b := g.resyncChan(), time.After(100*time.Millisecond)
+	a, _ := g.resyncChan()
+	b := time.After(100 * time.Millisecond)
 	select {
 	case <-a:
 		t.Logf("got timeout as expected")
@@ -104,6 +106,18 @@ func TestReflector_resyncChan(t *testing.T) {
 	}
 }
 
+func BenchmarkReflector_resyncChanMany(b *testing.B) {
+	s := NewStore(MetaNamespaceKeyFunc)
+	g := NewReflector(&testLW{}, &api.Pod{}, s, 25*time.Millisecond)
+	// The improvement to this (calling the timer's Stop() method) makes
+	// this benchmark about 40% faster.
+	for i := 0; i < b.N; i++ {
+		g.resyncPeriod = time.Duration(rand.Float64() * float64(time.Millisecond) * 25)
+		_, stop := g.resyncChan()
+		stop()
+	}
+}
+
 func TestReflector_watchHandlerError(t *testing.T) {
 	s := NewStore(MetaNamespaceKeyFunc)
 	g := NewReflector(&testLW{}, &api.Pod{}, s, 0)
diff --git a/pkg/controller/replication_controller.go b/pkg/controller/replication_controller.go
index 9e390e2999..1efb02ca41 100644
--- a/pkg/controller/replication_controller.go
+++ b/pkg/controller/replication_controller.go
@@ -57,7 +57,7 @@ const (
 	// specifically targeted at the case where some problem prevents an update
 	// of expectations, without it the RC could stay asleep forever. This should
 	// be set based on the expected latency of watch events.
-
+	//
 	// TODO: Set this per expectation, based on its size.
 	// Currently an rc can service (create *and* observe the watch events for said
 	// creation) about 10-20 pods a second, so it takes about 3.5 min to service
diff --git a/pkg/master/publish.go b/pkg/master/publish.go
index 3eebffdea4..dd3b48d04f 100644
--- a/pkg/master/publish.go
+++ b/pkg/master/publish.go
@@ -29,6 +29,8 @@ import (
 )
 
 func (m *Master) serviceWriterLoop(stop chan struct{}) {
+	t := time.NewTicker(10 * time.Second)
+	defer t.Stop()
 	for {
 		// Update service & endpoint records.
 		// TODO: when it becomes possible to change this stuff,
@@ -49,12 +51,14 @@ func (m *Master) serviceWriterLoop(stop chan struct{}) {
 		select {
 		case <-stop:
 			return
-		case <-time.After(10 * time.Second):
+		case <-t.C:
 		}
 	}
 }
 
 func (m *Master) roServiceWriterLoop(stop chan struct{}) {
+	t := time.NewTicker(10 * time.Second)
+	defer t.Stop()
 	for {
 		// Update service & endpoint records.
 		// TODO: when it becomes possible to change this stuff,
@@ -74,7 +78,7 @@ func (m *Master) roServiceWriterLoop(stop chan struct{}) {
 		select {
 		case <-stop:
 			return
-		case <-time.After(10 * time.Second):
+		case <-t.C:
 		}
 	}
 }
diff --git a/pkg/proxy/proxier.go b/pkg/proxy/proxier.go
index 5289e8a309..5587769fd0 100644
--- a/pkg/proxy/proxier.go
+++ b/pkg/proxy/proxier.go
@@ -366,16 +366,16 @@ const syncInterval = 5 * time.Second
 
 // SyncLoop runs periodic work.  This is expected to run as a goroutine or as the main loop of the app.  It does not return.
 func (proxier *Proxier) SyncLoop() {
+	t := time.NewTicker(syncInterval)
+	defer t.Stop()
 	for {
-		select {
-		case <-time.After(syncInterval):
-			glog.V(3).Infof("Periodic sync")
-			if err := iptablesInit(proxier.iptables); err != nil {
-				glog.Errorf("Failed to ensure iptables: %v", err)
-			}
-			proxier.ensurePortals()
-			proxier.cleanupStaleStickySessions()
+		<-t.C
+		glog.V(3).Infof("Periodic sync")
+		if err := iptablesInit(proxier.iptables); err != nil {
+			glog.Errorf("Failed to ensure iptables: %v", err)
 		}
+		proxier.ensurePortals()
+		proxier.cleanupStaleStickySessions()
 	}
 }
 
diff --git a/pkg/util/wait/wait.go b/pkg/util/wait/wait.go
index 1b05efe7d8..d6585cbe64 100644
--- a/pkg/util/wait/wait.go
+++ b/pkg/util/wait/wait.go
@@ -89,7 +89,12 @@ func poller(interval, timeout time.Duration) WaitFunc {
 			defer tick.Stop()
 			var after <-chan time.Time
 			if timeout != 0 {
-				after = time.After(timeout)
+				// time.After is more convenient, but it
+				// potentially leaves timers around much longer
+				// than necessary if we exit early.
+				timer := time.NewTimer(timeout)
+				after = timer.C
+				defer timer.Stop()
 			}
 			for {
 				select {