Well the only UNIX box I can think of which was able of parallele cmds was IBM SP (big blue...and followers).
I tried rdist and fell on your situation, and that was due to the loop...
Havent tried with rsync though... and havent tried multiple instance of expect, but did try at the time a workaround to speed up things:
schedule cron for the presence of a dummy file ( I did this for Y2K testings for shutdown of 30 servers.. )to execute your shutdown script.

---------- Post updated at 09:53 ---------- Previous update was at 09:48 ----------

Thinking of it now, you could try to schedule with at in your expect, that would be faster since after the at cmd (something like at now + 2 minutes...) it would have finished (if it works..) keep us informed (could be interesting!)