Yep, this is true, but I spent years working on AS/400 systems dealing with this exact issue. The thing is that the data isn't stored on a standard file-system. The record definition can be updated, in most situations, without modifying the application code very much at all. This isn't like an app that most people these days have experience with. IBM did everything they could so, if you followed the guidelines when the code was written, it was minimally painful to fix stuff like this in the future.
Everyone (including me) loves to give IBM shit, but they do actually know what they're doing when it comes to this stuff. There are many lessons and examples from mid-range and mainframe servers from IBM that are either ignored or re-invented in modern tech.