It is known that pyReScene tries to find out the correct rar version when reconstructing compressed rars by trial and error; the code for this is experimental. I would like to suggest to discuss possible improvements to this basic trial and error approach.
One straightforward possibility would be to go through the process of trial and error at the time of creating an .srr file, find out (if possible) the rar version used for compression by checking in a repository containing rar executables of all different versions available at that time, and adding this tiny (optional) information to the .srr file. Whenever the correct version can't be found at the time an .srr is created one would fallback to the trail and error process later when needed. This needs more time and computational power when creating .srr files, but once it is done every user can benefit from it later if the used rar version is found and incorporated into the .srr file.
Would it be possible to add an optional field to the .srr format to store a rar version used during compression? What other options do we have to improve the problem of discovery of rar versions?