Glad you asked Frick,
I'm still gathering more data points. I can't monopolize the machine, so I only get one test run every couple of days at best. And unfortunately, I just don't have any actual samples from the middle batch to compare to, just the original batch, and the latest one.
I ripped open one original sample and compared the batt cell by length to the ones from the assembly line. There is a surprising variability in cell lengths, up to 5mm, but the range is between 50-55mms. Which is consistent to my original sample, at 53mm.
I also had puff counts run between one unused 65mm and one 78mm batt from each batch, and the results have been consistent between the original and new, at 300-350 2-second puffs from the 65mm and 450-500 from the 78mm. And in fact the 65mm from the latest batch has been coming in consistently ahead of the old one.
Again, I did not have a sample from the suspect batch to test. There does seem to be up to a 10% variation in length normally. But that's not as much as you've observed, and that's just external length anyway. But I can say for sure that the latest batch matches up perfectly to the original.