I don’t think this relates to the problem.
The question rather is why TOPS Split has to be performed for every S1 image, even if all three swaths are required. Is it simply a matter of reducing the data size or maybe it has radiometric reasons.
The help of the TOPS Merge operator says:
“With the current toolbox, most operators in the TOPS InSAR processing chain operate on split product.”
To me that sounds like it is not entirely necessary but was initially designed to so. But maybe one of the developers can clarify.
By now, you have to split each swath and coregister them separately before you can use Back Geocoding on the result.
Source: Coregistrating more than two Sentinel-1 SLC IW products - #11 by jamesemwheeler