We have a couple of sources consisting of stylised parameters across a bunch of scenarios

$ ls -l /p/projects/rd3mod/inputdata/sources/industry_subsectors_specific/
total 66
-rw-rw-r-- 2 pehl rdev  494 Apr 21  2022 industry_specific_FE_limits.csv
-rw-rw-r-- 1 pehl rdev 4054 Aug 19  2022 specific_FE.csv
-rw-rw-r-- 1 pehl rdev 1017 Jun 17  2022 specific_material_alpha.csv
-rw-rw-r-- 1 pehl rdev  375 Mar 20  2023 specific_material_relative_change.csv
-rw-rw-r-- 1 pehl rdev 6021 Aug 19  2022 specific_material_relative.csv

It is entirely possible that these parameters are modified

1. without changing the file size, if a n-digit numbers are simply replaced by other n-digit numbers, or if the changes happen to add and remove the same number of characters, and
2. without changing anything in the first 300 bytes of the files, which in the case of industry_specific_FE_limits.csv are consumed entirely by the comment header and part of the csv header:

   $ head -c 300 industry_specific_FE_limits.csv 
   # Thermodynamic limits on industry specific FE consumption by Silvia Madeddu
   # (see post https://mattermost.pik-potsdam.de/rd3/pl/u7eg6ed5gpr85rabznepnaoqrr
   # and https://mattermost.pik-potsdam.de/rd3/pl/g74og14a7igi8n6trjbhgcntrc).
   # GJ/t for absolute subsectors, share for relative subsectors
   subse
   

   and in the case of specific_material_relative.csv make up just of five of the 136 data lines.

Analogous to this, if a file .fullhash is present in the directory, the entire files will be hashed, not just the first 300 bytes.