JDE runaway background write()? (bug?)

This is a bit of a long post about a bizarre bug I am experiencing in
my freshly-installed JDE.

I just installed JDE 2.3.5.1 on GNU Emacs 22.0.95.1
(x86_64-unknown-linux-gnu, GTK+ Version 2.4.13) of 2007-03-13. I
downloaded & installed CEDET version 1.0pre3 and elib 1.0.

I added the following to my .emacs:

    ;;; jdee
    (let ((jde-file (expand-file-name "~/.emacs.d/site-lisp/jde/lisp/jde"))
          (cedet-file (expand-file-name
"~/.emacs.d/site-lisp/cedet/common/cedet.el")))
      (when (and jde-file cedet-file)
        (log-feature 'jdee)
        ;; load CEDET
        (load-file cedet-file)
        ;; autoload jde-mode for .java files.
        (autoload 'jde-mode jde-file "JDE mode." t)
        (setq auto-mode-alist
              (append '(("\\.java\\'" . jde-mode)) auto-mode-alist))))

Visiting a .java file correctly runs JDE mode and displays "JDE
S/n/plugin/jdb Abbrev hs" in the mode line.

However, when JDE mode is active (i.e. when there is at least one
buffer my current buffers that is a jde-mode buffer) 100% of my CPU is
consumed by emacs. I investigated by strace(1)ing emacs, visiting a
JDE mode buffer, noting the CPU spike, then Ctrl-C'ing emacs, so that
the last thing in the strace log is whatever's going on in emacs.

The last lines of strace output read:

    write(4, ";\0\5\0\366\1\300\0\0\0\0\0\33\0\264\0X\2\f\0C\0\5\0\201"...,
56) = 56

and there are literally thousands of these lines in the strace output.

Since it was writing (consistently across runs) to fd 4, I thought I'd
run lsof(8) and grep for the PID of the emacs instance running JDE,
and see what fd 4 corresponds to. It said:

,----lsof output
| emacs     16120 db276    4u     unix 0x000001007ca40c80
 286001 socket
`----

Interestingly, when I try to type in an emacs buffer, the CPU usage
darts back down to normal. I can type at normal speed in a JDE buffer
just fine. When I'm *not* typing, however, is when this occurs.

_Summary_
For some reason, JDE is loop-writing to a unix socket as fast as
possible, when the user is doing nothing.

I am currently looking at other writes to fd 4 in the strace output,
to see if I can discern what kind of data is being stuffed into that
file descriptor.

-Denis