diff --git a/go/vt/discovery/healthcheck.go b/go/vt/discovery/healthcheck.go
index 7fde2774bd4..8a840dd3471 100644
--- a/go/vt/discovery/healthcheck.go
+++ b/go/vt/discovery/healthcheck.go
@@ -527,24 +527,17 @@ func (hc *HealthCheckImpl) checkConn(hcc *healthCheckConn, name string) {
 		// between the goroutine that sets it and the check for its value
 		// later.
 		timedout := sync2.NewAtomicBool(false)
-		serving := hcc.tabletStats.Serving
 		go func() {
 			for {
 				select {
-				case serving = <-servingStatus:
+				case <-servingStatus:
 					continue
 				case <-time.After(hc.healthCheckTimeout):
-					// Ignore if not serving.
-					if !serving {
-						continue
-					}
 					timedout.Set(true)
 					streamCancel()
 					return
 				case <-streamCtx.Done():
-					// If stream returns while serving is false, the function
-					// will get stuck in an infinite loop. This code path
-					// breaks the loop.
+					// If the stream is done, stop watching.
 					return
 				}
 			}
diff --git a/go/vt/discovery/healthcheck_test.go b/go/vt/discovery/healthcheck_test.go
index 980eb90cadd..caa61e2d1fd 100644
--- a/go/vt/discovery/healthcheck_test.go
+++ b/go/vt/discovery/healthcheck_test.go
@@ -545,28 +545,31 @@ func TestHealthCheckTimeout(t *testing.T) {
 		t.Errorf("StreamHealth should be canceled after timeout, but is not")
 	}
 
-	// repeat the wait. There should be no error or cancelation.
+	// repeat the wait. It will timeout one more time trying to get the connection.
 	fc.resetCanceledFlag()
-	time.Sleep(2 * timeout)
+	time.Sleep(timeout)
 	t.Logf(`Sleep(2 * timeout)`)
 
-	select {
-	case res = <-l.output:
-		t.Errorf(`<-l.output: %+v; want not message`, res)
-	default:
+	res = <-l.output
+	if res.Serving {
+		t.Errorf(`<-l.output: %+v; want not serving`, res)
 	}
 
-	if err := checkErrorCounter("k", "s", topodatapb.TabletType_MASTER, 1); err != nil {
+	if err := checkErrorCounter("k", "s", topodatapb.TabletType_MASTER, 2); err != nil {
 		t.Errorf("%v", err)
 	}
 
-	if fc.isCanceled() {
-		t.Errorf("StreamHealth should not be canceled after timeout")
+	if !fc.isCanceled() {
+		t.Errorf("StreamHealth should be canceled again after timeout")
 	}
 
 	// send a healthcheck response, it should be serving again
+	fc.resetCanceledFlag()
 	input <- shr
 	t.Logf(`input <- {{Keyspace: "k", Shard: "s", TabletType: MASTER}, Serving: true, TabletExternallyReparentedTimestamp: 10, {SecondsBehindMaster: 1, CpuUsage: 0.2}}`)
+
+	// wait for the exponential backoff to wear off and health monitoring to resume.
+	time.Sleep(timeout)
 	res = <-l.output
 	if !reflect.DeepEqual(res, want) {
 		t.Errorf(`<-l.output: %+v; want %+v`, res, want)