I'll try a 1.8x branch next to see if we can replicate there (and, if necessary, will push trunk against 2.9x to see if I can make that flake). Will update with results after I get it going.

Test server is my local VM :). Took a bit to get it to compile, but a few tweaks to POM version #s and a bunch of 'singleton="true"'s later it was working well enough to test with. I didn't do any extended validation, but I wasn't noticing anything out of the ordinary in quick testing either.

Your attention to the issue is very much appreciated.... this may be a useful patch for 1.8.x, perhaps? I just started working on getting GB2 to compile nicely with 10 and will prob setup a maven profile for it. Where is the test server with 10 and gb2?

Finally managed to get on the 10 test server, and I agree - my test case fails on this release. Will see if I can track down what changed in the duplication code that the 'null's aren't persisted (if anyone has suggestions, I'm all eyes), suffice the test case doesn't apply to trunk, and this should probably just be a minor / 'nice to have' sort of bug rather than how I originally classified. Sorry 'bout that!

Best guess is that it's something which was changed in the site replication code - I can only test against Sak 2.9.x / GB2 1.8.x right now, but will try to get on a .10x / trunk build tomorrow to see if I can replicate there. Suffice, I think the fix stands independent as NPE-protective and more consistent with the surrounding code, but I'll try to get a better test procedure in place which won't just work on my local prod box ;).

Thanks for adding detail. Unfortunately it still didn't fail. For me.. one thing I noticed is that the item was already 'included in the grade' (checkbox checked)... I tried uncheck with no error, but the polarity was different than what your #5 made me think. So I wonder what the diff was