Well, it looks like I got lucky and found the cause pretty quickly. Since the   
   problems were so minor and seemed like they might be timing related, I started   
   with the CIA chips. I happen to have a working spare, and tried that in both   
   sockets.   
      
   After I replaced the 2nd CIA (U2), the problems were gone. 3 runs of the demo   
   with the new CIA, not a single glitch. Then I double checked by re-installing   
   the old chip. 3 runs, glitches every time. Re-installed the new chip, 3 runs   
   without problems.   
      
   I'm not sure if there is a tiny fault with the old CIA chip, or if there are   
   subtle timing differences between different chips that could cause this. Both   
   the old chips are 6526, but the 2nd is taken from an NTSC machine. Could that   
   make a difference, or    
   are they exactly the same in PAL/NTSC?   
      
   The new one is unfortunately 6526A, which from what I understand can cause   
   minor incompatibilities with some demos. I guess I'll have to test and see   
   which chip works better for me.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)