Hmmm... Generally we experienced quite a lot of "strangeness" with the iiwa over the years. We had a process where the tool was vibrating and the robot stopped reliably after (I don't exactly remember) like 30 seconds, because some internal timeout happened. When we did a polishing project for a company, we tested all kinds of movement types and had a lot of problems with very dense spacing of positions, where the robot would stop in-between or even crash. There were then also changes depending on software-version etc.
Unfortunately that shows that the Sunrise platform is just not as tested and reliable as let's say KSS.
Regarding your specific error I would guess that some internal system crashes and the safety picks up on that and stops the robot. By default, commands are grouped into MotionBatches. You could look into the Java file and change the number of commands per Motion Batch, that could maybe help.
Regarding the different files: Have you tried running the same definition on both machines and compare the output? I can also do that from my side, if you attach both XML files.
If it is the same KUKA|prc and Rhino version, I don't really see what could cause that. We had some issues with different language settings (e.g. in German you would express 4.5 as 4,5) but such an error would be very obvious and not only happen sometimes.
Sorry that I don't have a specific solution right now, but I'd be happy to look at XML files from two different PCs and check whether they are identical.