I can't really say much about the lack of responsiveness and the double
submits, but I can make a couple of suggestions about some of your other
issues.  (Sorry for the delay -- I'm off-campus right now.)

- If you take a look at the Swiki Swiki
(http://pbl.cc.gatech.edu:8080/myswiki.1) you'll find some suggestions for
how to make an image start up with Squeak *without* tweaking the startUp
method.  My favorite (contributed by Andreas, I believe) is to put all the
restore and startup code in a workspace, then put at the top "Smalltalk
snapshot: true andQuit: true."  Select the whole thing and do-it.  The
image saves while in the midst of the do-it.  When the image starts back
up, the restore and server startup code is processed.  Works great!

- Have you tried Commanche yet?  It uses the ConnectionQueue differently.
In Bolot and Jochen's tests so-far, it outperforms PWS in terms of
responsiveness and fewer errors.  The test Commanche Swiki is at
http://seaweed.cc.gatech.edu:8080/refs/1

- I wonder if we're running into a problem with being midway in the
implementation of the accept() socket semantics?  That it's in the image,
but not in (all? any?) of the VMs?

Good luck!
  Mark

--------------------------
Mark Guzdial : Georgia Tech : College of Computing : Atlanta, GA 30332-0280
(404) 894-5618 : Fax (404) 894-0673 : guzdial at cc.gatech.edu
http://www.cc.gatech.edu/gvu/people/Faculty/Mark.Guzdial.html