If they are biological replicates, I'd recommend running ROSE on each separately and then comparing overlap and signal with bedtools/diffbind/csaw. When comparing SEs, signal changes over such large windows often dilutes the fold changes that you'll get from most approaches. In my opinion, the best way to go about it is to compare the constituent elements of each super enhancer between groups.

In short, coming up with SEs for each sample, potentially merging them to derive a consensus set, and then label all of the constituent peaks so you know they lie in an SE. Looking at differing signal at the constituent peaks doesn't generally run into the same issues that comparing the entire SE directly does statistically.