Please try using this CVS snapshot:

  http://snaps.php.net/php4-STABLE-latest.tar.gz
 
For Windows:
 
  http://snaps.php.net/win32/php4-win32-STABLE-latest.zip


..and you'll know for sure if it's related to #21529 since
that was fixed in CVS..

Nope, problem still exists.  As of right now, it's hanging on a piece of code that looks like this:

------------ CODE ------------

print "Unresponsive server, closing connection... ";
ob_flush();

if (is_resource($this->m_rConn))
  fclose($this->m_rConn);

$this->m_rConn = NULL;
print "Done!\n";
ob_flush();

------------ CODE ------------

Note that it's not printing "Done!" in the spot where it is freezing.  Meaning it's either is_resource, or fclose.  How's that for nasty?

Would it be possible to give us some complete (but short!) 
example script that we can use to test this ourselves? 

You've got some objects mixed there and I guess you're
also using some output buffering too?

If possible, try creating a script that only has the socket
stuff in it, without the ob_* and OO.

Well, originally I had the ob_flush in there because flush and ob_implicit_flush didn't work, but that seems to be fixed now.  Making this test script was *not* easy; tracking this bug down has been driving me nuts.  If you want my list of hosts, you're welcome to it, but it froze on mail.uu.net last time I ran it.  Here you go:

<?PHP

// Turn on flushing so we can see debug output.

ob_implicit_flush(1);

// hosts.txt is a file of about 170 hosts I'm sending mail
// to with my SMTP class.  Don't worry, I'm not a spammer,
// we run a home-grown mailing list software our newspapers
// use - strictly subscription based.

$aHosts = file('hosts.txt');

foreach ($aHosts as $sDomain)
{
  $sDomain = rtrim($sDomain);

  print "Testing $sDomain...\n";

  $aMX = array();

  // Obviously I skip hosts with no MX records.

  if (!getmxrr($sDomain, $aMX))
    continue;

  foreach ($aMX as $sMXHost)
  {
    print "Connecting to $sMXHost...\n";

    $nError = $sError = NULL;
    $rSock = fsockopen($sMXHost, 25, $nError, $sError, 2);

    if (!is_resource($rSock))
    {
      print "Error '$sError' skipping.\n";
      continue;
    }
    stream_set_timeout($rSock, 2);

    // Now we read to EOF, since that seems to be where
    // the problem is.  I'm also putting in a PHP timeout
    // to make sure the while loop ends.

    $nTimeout = time() + 3;

    while (is_resource($rSock) && !feof($rSock) &&
           time() < $nTimeout)
    {
      print fgets($rSock, 1024);
    }

    // And here's the issue.  Eventually the script will just
    // freeze, before it gets to the last print statement.

    print "Closing connection...";
    if (is_resource($rSock))
      fclose($rSock);
    print "Done!\n";
  }
}

?>

And here's a new backtrace for the php4-STABLE tarball I downloaded last night.

#0  0x401c941e in select () from /lib/libc.so.6
#1  0x08199a44 in __DTOR_END__ ()
#2  0x08122165 in _php_stream_free (stream=0x81eb54c, close_options=3)
    at /home/install/php4-STABLE-200302062230/main/streams.c:327
#3  0x080ad01f in zif_fclose (ht=1, return_value=0x81e7c2c, this_ptr=0x0, 
    return_value_used=0)
    at /home/install/php4-STABLE-200302062230/ext/standard/file.c:1120
#4  0x0815b2d0 in execute (op_array=0x81d8c44)
    at /home/install/php4-STABLE-200302062230/Zend/zend_execute.c:1596
#5  0x0814560c in zend_execute_scripts (type=8, retval=0x0, file_count=3)
    at /home/install/php4-STABLE-200302062230/Zend/zend.c:864
#6  0x08118b39 in php_execute_script (primary_file=0xbffffa50)
    at /home/install/php4-STABLE-200302062230/main/main.c:1582
#7  0x08164180 in main (argc=2, argv=0xbffffaf4)
    at /home/install/php4-STABLE-200302062230/sapi/cli/php_cli.c:753
#8  0x4010a1c4 in __libc_start_main () from /lib/libc.so.6

Duplicate of #21809.

This bug has been fixed in CVS.

In case this was a PHP problem, snapshots of the sources are packaged
every three hours; this change will be in the next snapshot. You can
grab the snapshot at http://snaps.php.net/.
 
In case this was a documentation problem, the fix will show up soon at
http://www.php.net/manual/.

In case this was a PHP.net website problem, the change will show
up on the PHP.net site and on the mirror sites in short time.
 
Thank you for the report, and for helping us make PHP better.