The -scale_invnodevol option performs a scaling quite different to that of -scale_invlength: it is based on the volumes of the two parcels involved in the connection, and is not related to the length of the connection. This scaling mechanism is also described in the manuscript mentioned above.
I wouldn’t say that either of these weights are “better” at “normalizing” the connection strengths. The effectiveness of -scale_invlength relative to the SIFT method is discussed in this manuscript. I also rambled about these issues in this blog post. While inter-hemispheric connections may be relatively sparse (~ 2% of all white matter fibres, as I mentioned in this article), one must be extremely careful about understanding how such observations are made, whether they are in fact “biases” that can / should be “fixed”, and the potential ramifications of any heuristic mechanisms applied.
Rob