Mostly bank specific: the parsing and rounding approach wasn’t always consistent depending on what was being run. Still crazy to me that it was all CSV (with some special extra formatting/structure) via FTP to run transactions too at the time!