Yes, I had missed the beginning of this conversation when I reposted earlier today as an updated comment from a past incident. In that case, David Logan was able to solve the issue with some kind of reformatting of the figures, but this was definitely a work-around and they were able to fix it in an update.
I am pretty sure the numbers change because CP can identify the primary objects for an image but then forgets how to count them (beginning with 0 instead of 1), so they are different for every image. So in the case you cite in the beginning, the 227 objects it found are numbered from 0 to 226.
In my case, the problem is only for some images, but it consistently has problems with these images. If I remove them, the pipeline will run without an error, but it can be a problem for a third or more of the images. The images themselves don’t appear different to me from those that don’t cause problems.
I am using 3.1.5 on a Mac. Does anyone know of a build that does not have this problem? It would be best for me to stick with CP3, but I would go back to CP2 if needed.