Hello :-)
Have you considered fine-graining the analysis to investigate how adding one of the above-mentioned files to a software's source code affects citations? I started compiling a (still very) small dataset of Pull/Merge requests with such files. The GitHub & GitLab APIs could be used to get the merge-date, next release version etc.. Then, one might be able to find out, whether or not there was an uptick in citations after such software citation approach was used, via a time-series analysis (or something similar; I'm a novice there).
Cheers!
PS: Possibly related to force11/force11-sciwg#2 & force11/force11-sciwg#4