I’ve discovered a fantastic new feature in Siril: you can now have it apply background extraction per frame in a sequence.
Since the sky gradient can be quite different between sub frames in a project, especially when tracked all throughout the sky over multiple nights, stacking all of the frames can create a very complex background gradient in your final integration. At least in Siril, the background extraction process is limited to fitting polynomials of up to the fourth degree. The gradient won’t typically be a close fit from any polynomial, so background correction can be quite limited. This limits how much you can stretch the background to expose dim details.
As an example, my last project, the Pinwheel Cluster, was drastically limited by gradients from light pollution and the moon. There is nebulosity in the background of this data, but my initial processing lost it due to very bad gradients that needed to just be black clipped. I decided to try integrating the data again, but a second order background extraction was applied to each sub between the registration and stacking processes. This led to a much, much cleaner background in the final stack, and I could stretch the final image a lot more to reveal dim nebulosity.