No, the buffer size didn't change. I think the expected tab can sometimes consume all of the available space (even though I haven't encountered this myself) since its content is not truncated to the first 100 rows like it happens for the actual tab. Anyway, I replaced the CW-provided testing function with the custom one I use, which I somehow forgot to do while updating all of my SQL katas earlier, so this shouldn't happen anymore if your solution is fully correct; otherwise, it'll still output the whole table, both on actual and expected tabs.