>Retorquing doesn't imply tightening beyond specs. It means retightening 
>TO specs. I think the logic goes that vibration and heat allow the 
>torqued bolts to settle some...thus losing the to-spec torque that you 
>provided originally. Retorquing puts everything back to where it ought to 
>be...sort of like that disclaimer on the cereal box: some settling of 
>contents may occur.
>

YES

>> Does it hold ignition compression? Or is it a guard against
>> a fatigue failure of something or the other?
>>  

YES

The long studs that hold your heads onto the barrels are subject to fatigue 
loading from the firing.  ANY bolted connection is much less susceptible to 
fatigue(cyclic) failure if it has a high preload on it.  This seems counter 
productive but... This raises the steady state stress level such that the 
cyclic stresses become a smaller percentage of total load.  This keeps 
fatigue away, not so much in the studs as in the thread roots in the case! 
Time to stop lecturing.  Terms (almost) over.
           Oh, I try to tighten both sides of the engine up to spec at the 
same time, slowly raise it up, then do the VW two-step simultaneously to set 
the final torques. I have a theory, this theory which I have is mine and my 
theory is that you can warp the case if you tighten up one side to 100% 
while the other is loose.  Am I nuts or what? I said it was a theory! Figure 
it's the cause of case leaks along the seam.   tim