diff --git a/util/sync.go b/util/sync.go
index 9d5a66ee..0ceeebe0 100644
--- a/util/sync.go
+++ b/util/sync.go
@@ -229,3 +229,82 @@ func (obj *SubscribedSignal) Send() {
 
 	// release (re-use the above mutex)
 }
+
+// BoundedReadSemaphore is a mutex that allows multiple Lock operations to occur
+// concurrently, as if they were read locks. The distinction is that for the
+// first Lock operation to complete, the Start() channel state must be read. At
+// this point subsequent Lock operations will succeed. The End state completes
+// once the last paired Unlock operation is run. The cycle can be repeated
+// without needing to re-initialize the struct. Each Lock or Unlock operation
+// itself contains a call to Lock an internal mutex for accounting and
+// implementation purposes.
+//
+// This was previously named SharedMutex. We welcome alternate naming
+// suggestions.
+type BoundedReadSemaphore struct {
+	// mutex locks individual operations on our struct.
+	mutex *sync.Mutex
+
+	// start is the Start signal channel.
+	start chan struct{}
+
+	// end is the End signal channel.
+	end chan struct{}
+
+	// count keeps track of the number of active lockers.
+	count int64
+}
+
+// NewBoundedReadSemaphore creates an initialized object. This must be done
+// before first use, as the empty struct is not a valid BoundedReadSemaphore.
+func NewBoundedReadSemaphore() *BoundedReadSemaphore {
+	return &BoundedReadSemaphore{
+		mutex: &sync.Mutex{},
+		start: make(chan struct{}),
+		end:   make(chan struct{}),
+	}
+}
+
+// Lock asks for a lock on this mutex. After the first Lock call synchronizes,
+// subsequent calls will succeed quickly. The first "synchronization" waits for
+// the lock start signal, which is a receive on the Start channel.
+func (obj *BoundedReadSemaphore) Lock() {
+	obj.mutex.Lock()
+	defer obj.mutex.Unlock()
+
+	obj.count++
+
+	if obj.count == 1 { // we're the first lock (the mutex guarantees this)
+		obj.start <- struct{}{} // lock start
+	}
+}
+
+// Unlock unlocks from a previous Lock operation. If this call is the last
+// paired unlock operation, it blocks until the End() signal is synchronized. If
+// you unlock more times than you lock, then you will cause a panic.
+func (obj *BoundedReadSemaphore) Unlock() {
+	obj.mutex.Lock()
+	defer obj.mutex.Unlock()
+
+	if obj.count == 0 {
+		panic("negative BoundedReadSemaphore counter")
+	}
+
+	obj.count--
+
+	if obj.count == 0 { // we're the last unlock (the mutex guarantees this)
+		obj.end <- struct{}{} // lock end
+	}
+}
+
+// Start will return a single signal on this output channel to advise that we've
+// started the "locked" state.
+func (obj *BoundedReadSemaphore) Start() <-chan struct{} {
+	return obj.start
+}
+
+// End will return a single signal on this output channel to advise that we've
+// returned to the "unlocked" state.
+func (obj *BoundedReadSemaphore) End() <-chan struct{} {
+	return obj.end
+}
diff --git a/util/sync_test.go b/util/sync_test.go
index e00022bf..d594ae90 100644
--- a/util/sync_test.go
+++ b/util/sync_test.go
@@ -167,3 +167,107 @@ func ExampleSubscribedSignal() {
 	// done sending signal
 	// exiting...
 }
+
+func ExampleBoundedReadSemaphore() {
+	fmt.Printf("hello\n")
+	defer fmt.Printf("goodbye\n")
+
+	wg := &sync.WaitGroup{}
+	defer wg.Wait()
+
+	ch := make(chan struct{}) // close signal
+
+	brs := NewBoundedReadSemaphore()
+
+	wg.Add(1)
+	go func() {
+		defer wg.Done()
+		brs.Lock()
+		defer brs.Unlock()
+		time.Sleep(100 * time.Millisecond) // delay for consistent print
+
+		fmt.Printf("#1 is in the locked zone\n")
+		time.Sleep(1 * time.Second)
+	}()
+
+	wg.Add(1)
+	go func() {
+		defer wg.Done()
+		brs.Lock()
+		defer brs.Unlock()
+		time.Sleep(200 * time.Millisecond) // delay for consistent print
+
+		fmt.Printf("#2 is in the locked zone\n")
+		time.Sleep(2 * time.Second)
+	}()
+
+	wg.Add(1)
+	go func() {
+		defer wg.Done()
+		brs.Lock()
+		defer brs.Unlock()
+		time.Sleep(300 * time.Millisecond) // delay for consistent print
+
+		fmt.Printf("#3 is in the locked zone\n")
+		time.Sleep(3 * time.Second)
+	}()
+
+	wg.Add(1)
+	go func() {
+		defer wg.Done()
+		defer close(ch) // exit signal
+		max := 2        // configure me
+		for {
+			if max == 0 {
+				break
+			}
+			max--
+			time.Sleep(4 * time.Second)
+			brs.Lock()
+			time.Sleep(100 * time.Millisecond) // delay for consistent print
+			fmt.Printf("#4 is in the locked zone\n")
+
+			brs.Unlock()
+			time.Sleep(100 * time.Millisecond) // delay for consistent print
+			fmt.Printf("#4 is in the unlocked zone\n")
+		}
+	}()
+
+Loop:
+	for {
+		select {
+		case <-ch: // exit signal
+			break Loop
+
+		case <-brs.Start(): // An empty value is received to start the locking.
+			fmt.Printf("shared mutex start\n")
+		}
+
+		// subsequent Lock's that happen when at least one Lock is
+		// already held are permitted...
+		time.Sleep(1 * time.Second)
+
+		// something happens here
+
+		select {
+		case <-brs.End(): // An empty values is received when the last Unlock happens.
+			fmt.Printf("shared mutex end\n")
+		}
+	}
+
+	// Output: hello
+	// shared mutex start
+	// #1 is in the locked zone
+	// #2 is in the locked zone
+	// #3 is in the locked zone
+	// shared mutex end
+	// shared mutex start
+	// #4 is in the locked zone
+	// shared mutex end
+	// #4 is in the unlocked zone
+	// shared mutex start
+	// #4 is in the locked zone
+	// shared mutex end
+	// #4 is in the unlocked zone
+	// goodbye
+}