The basic method behind making such a determination is that you partition the segments into subsets and look at precede-and-follow restrictions in the language, and impute those restrictions to the notion of "sonority hierarchy". This is based on the premise that within the syllable, sonority increases up to the peak, then decreases, and ordering is based on sonority. For example, syllables in English start with [pl] but not [lp]; then end with [lp] and not [pl]. Therefore, [p] is less sonorous than [l]. Since syllables can end [jl] and not [lj] we surmise that [j] is more sonorous than [l]. In addition, there are some a priori concepts about sonority, based on the idea that "sonorous" means "more vowe-like", so that [p] is very un-vowellike and [m] is relatively more vowel-like. However, many languages have syllable-initial NC clusters – known as a "sonority reversal". So you can't absolutely use sequencing as the determinant of sonority.
There is generally a tendency to put all vowels into a single sonority category (most sonorous), however, modern Greek vowel hiatus resolutions is "hierarchical" (in many dialects) and thus people actually say that dialects differ as to the sonority of particular vowels (Newton, Kaisse have written on this).
Steriade in her dissertation exploits sonority to account for syllabification in Greek, and Lehmann invokes sonority in Latin syllabification. Sonority is called on in a number of publications on Icelandic syllabification, which could substitute for Old Norse. This paper makes weak sonority claims about Russian.