I ended up stacking before and after the meridian flip separately (since the gradients were opposite). Then I applied the DBE tool in PixInsight to them separately to get rid of the gradients. Then I stacked the results together. Processing then proceeded as normal. It was more trouble but at least it did the trick.
This one reminds me a bit of a brain with the various folds. The embedded cluster in the middle left is Berkeley 90.