diff --git a/winsup/cygwin/ChangeLog b/winsup/cygwin/ChangeLog
index a02ca36a8..0ac85de9f 100644
--- a/winsup/cygwin/ChangeLog
+++ b/winsup/cygwin/ChangeLog
@@ -1,3 +1,8 @@
+2013-06-07  Christopher Faylor  <me.cygwin2013@cgf.cx>
+
+	* DevNotes: Add entry cgf-000023.
+	* sigproc.cc (exit_thread): Remove now-unneeded sleep code.
+
 2013-06-05  Corinna Vinschen  <corinna@vinschen.de>
 
 	* include/cygwin/version.h (CYGWIN_VERSION_DLL_MINOR): Bump to 21.
diff --git a/winsup/cygwin/DevNotes b/winsup/cygwin/DevNotes
index 70f4ec22a..3016075ec 100644
--- a/winsup/cygwin/DevNotes
+++ b/winsup/cygwin/DevNotes
@@ -1,3 +1,8 @@
+2013-06-07  cgf-000023
+
+Given the fact that the signal thread never exits there is no need
+for exit_thread to ever block.  So, nuke this code.
+
 2013-01-31  cgf-000022
 
 While researching the lftp behavior reported here:
diff --git a/winsup/cygwin/sigproc.cc b/winsup/cygwin/sigproc.cc
index 55c5bbac9..80a6fff1c 100644
--- a/winsup/cygwin/sigproc.cc
+++ b/winsup/cygwin/sigproc.cc
@@ -447,21 +447,6 @@ exit_thread (DWORD res)
 # undef ExitThread
   sigfillset (&_my_tls.sigmask);	/* No signals wanted */
   lock_process for_now;			/* May block indefinitely when exiting. */
-  /* ES_EXIT_STARTING indicates that exit is in progress.  After setting
-     exit_state to ES_EXIT_STARTING, the global dtors are running first,
-     then the exit state is set to the next level in do_exit.  We must not
-     block the thread exit while the global dtors are running, because
-     one of them might have called pthread_join, which is perfectly valid
-     for a global C++ destructor.
-     FIXME: Do we need another state between ES_EXIT_STARTING and
-            ES_SIGNAL_EXIT to narrow the gap in which the thread exit
-	    is still valid? */
-  if (exit_state > ES_EXIT_STARTING)
-    {
-      for_now.release ();
-      Sleep (INFINITE);
-    }
-
   HANDLE h;
   if (!DuplicateHandle (GetCurrentProcess (), GetCurrentThread (),
                         GetCurrentProcess (), &h,