From: Marc LeBrun <fxpt.com at mlb>

  The probability of two programs initializing [a random seed]
   within exactly the same window is probably less than lightning striking the 
   user.

Not so - it happened to me quite often.  I was starting batch Allegro
jobs on a dual-processor Sun, using a script.  The random seed was
inited early in the job, and often came out the same for both
processes.  My hypothesis was that the random seed was based on a
clock that got incremented each 1/60 second, and that both processes
had to wait for the Lisp image to arrive from the file server, then
got started within 1/60 second of each other.  This is only a
hypothesis, but it is a fact that the random seeds were sometimes the
same.