diff --git a/examples/src/main/java/io/grpc/examples/manualflowcontrol/ManualFlowControlServer.java b/examples/src/main/java/io/grpc/examples/manualflowcontrol/ManualFlowControlServer.java
index abd04fa9ad..694330dfdb 100644
--- a/examples/src/main/java/io/grpc/examples/manualflowcontrol/ManualFlowControlServer.java
+++ b/examples/src/main/java/io/grpc/examples/manualflowcontrol/ManualFlowControlServer.java
@@ -24,7 +24,6 @@ import io.grpc.stub.StreamObserver;
 
 import java.io.IOException;
 import java.util.concurrent.TimeUnit;
-import java.util.concurrent.atomic.AtomicBoolean;
 import java.util.logging.Logger;
 
 public class ManualFlowControlServer {
@@ -42,12 +41,6 @@ public class ManualFlowControlServer {
             (ServerCallStreamObserver<HelloReply>) responseObserver;
         serverCallStreamObserver.disableAutoInboundFlowControl();
 
-        // Guard against spurious onReady() calls caused by a race between onNext() and onReady(). If the transport
-        // toggles isReady() from false to true while onNext() is executing, but before onNext() checks isReady(),
-        // request(1) would be called twice - once by onNext() and once by the onReady() scheduled during onNext()'s
-        // execution.
-        final AtomicBoolean wasReady = new AtomicBoolean(false);
-
         // Set up a back-pressure-aware consumer for the request stream. The onReadyHandler will be invoked
         // when the consuming side has enough buffer space to receive more messages.
         //
@@ -55,10 +48,17 @@ public class ManualFlowControlServer {
         // onNext(), onError(), and onComplete() handlers. Blocking the onReadyHandler will prevent additional messages
         // from being processed by the incoming StreamObserver. The onReadyHandler must return in a timely manor or else
         // message processing throughput will suffer.
-        serverCallStreamObserver.setOnReadyHandler(new Runnable() {
+        class OnReadyHandler implements Runnable {
+          // Guard against spurious onReady() calls caused by a race between onNext() and onReady(). If the transport
+          // toggles isReady() from false to true while onNext() is executing, but before onNext() checks isReady(),
+          // request(1) would be called twice - once by onNext() and once by the onReady() scheduled during onNext()'s
+          // execution.
+          private boolean wasReady = false;
+
           @Override
           public void run() {
-            if (serverCallStreamObserver.isReady() && wasReady.compareAndSet(false, true)) {
+            if (serverCallStreamObserver.isReady() && !wasReady) {
+              wasReady = true;
               logger.info("READY");
               // Signal the request sender to send one message. This happens when isReady() turns true, signaling that
               // the receive buffer has enough free space to receive more messages. Calling request() serves to prime
@@ -66,7 +66,9 @@ public class ManualFlowControlServer {
               serverCallStreamObserver.request(1);
             }
           }
-        });
+        }
+        final OnReadyHandler onReadyHandler = new OnReadyHandler();
+        serverCallStreamObserver.setOnReadyHandler(onReadyHandler);
 
         // Give gRPC a StreamObserver that can observe and process incoming requests.
         return new StreamObserver<HelloRequest>() {
@@ -90,16 +92,17 @@ public class ManualFlowControlServer {
               // Check the provided ServerCallStreamObserver to see if it is still ready to accept more messages.
               if (serverCallStreamObserver.isReady()) {
                 // Signal the sender to send another request. As long as isReady() stays true, the server will keep
-                // cycling through the loop of onNext() -> request()...onNext() -> request()... until either the client
-                // runs out of messages and ends the loop or the server runs out of receive buffer space.
+                // cycling through the loop of onNext() -> request(1)...onNext() -> request(1)... until the client runs
+                // out of messages and ends the loop (via onCompleted()).
                 //
-                // If the server runs out of buffer space, isReady() will turn false. When the receive buffer has
-                // sufficiently drained, isReady() will turn true, and the serverCallStreamObserver's onReadyHandler
-                // will be called to restart the message pump.
+                // If request() was called here with the argument of more than 1, the server might runs out of receive
+                // buffer space, and isReady() will turn false. When the receive buffer has sufficiently drained,
+                // isReady() will turn true, and the serverCallStreamObserver's onReadyHandler will be called to restart
+                // the message pump.
                 serverCallStreamObserver.request(1);
               } else {
                 // If not, note that back-pressure has begun.
-                wasReady.set(false);
+                onReadyHandler.wasReady = false;
               }
             } catch (Throwable throwable) {
               throwable.printStackTrace();
diff --git a/stub/src/main/java/io/grpc/stub/CallStreamObserver.java b/stub/src/main/java/io/grpc/stub/CallStreamObserver.java
index 8ed16bb96e..98fa6fba57 100644
--- a/stub/src/main/java/io/grpc/stub/CallStreamObserver.java
+++ b/stub/src/main/java/io/grpc/stub/CallStreamObserver.java
@@ -19,20 +19,30 @@ package io.grpc.stub;
 import io.grpc.ExperimentalApi;
 
 /**
- * A refinement of StreamObserver provided by the GRPC runtime to the application that allows for
- * more complex interactions with call behavior.
+ * A refinement of StreamObserver provided by the GRPC runtime to the application (the client or
+ * the server) that allows for more complex interactions with call behavior.
  *
- * <p>In any call there are logically two {@link StreamObserver} implementations:
+ * <p>In any call there are logically four {@link StreamObserver} implementations:
  * <ul>
- *   <li>'inbound' - which the GRPC runtime calls when it receives messages from the
- *   remote peer. This is implemented by the application.
+ *   <li>'inbound', client-side - which the GRPC runtime calls when it receives messages from
+ *   the server. This is implemented by the client application and passed into a service method
+ *   on a stub object.
  *   </li>
- *   <li>'outbound' - which the GRPC runtime provides to the application which it uses to
- *   send messages to the remote peer.
+ *   <li>'outbound', client-side - which the GRPC runtime provides to the client application and the
+ *   client uses this {@code StreamObserver} to send messages to the server.
+ *   </li>
+ *   <li>'inbound', server-side - which the GRPC runtime calls when it receives messages from
+ *   the client. This is implemented by the server application and returned from service
+ *   implementations of client-side streaming and bidirectional streaming methods.
+ *   </li>
+ *   <li>'outbound', server-side - which the GRPC runtime provides to the server application and
+ *   the server uses this {@code StreamObserver} to send messages (responses) to the client.
  *   </li>
  * </ul>
  *
- * <p>Implementations of this class represent the 'outbound' message stream.
+ * <p>Implementations of this class represent the 'outbound' message streams. The client-side
+ * one is {@link ClientCallStreamObserver} and the service-side one is
+ * {@link ServerCallStreamObserver}.
  *
  * <p>Like {@code StreamObserver}, implementations are not required to be thread-safe; if multiple
  * threads will be writing to an instance concurrently, the application must synchronize its calls.