I worked for a company years ago the had multi member files. There were members like PROD, TEST, ENG, DEV.
Not our decision, it was prepackaged software.
PROD was for production work, TEST for programmer code testing, DEV for the product development team and ENG for the product engineers.
This was so we could have data unique for each business sector. It prevented us from changing the data that did not belong to us, It was also a pain to code for. You had to do a lot of OVRDBF commands to point to the correct member to use. It would have been much easier to just have a duplicate file in another library that could be controlled by a JOBD or LIBL.
The other issue is if you have a system or file crash, it takes them all down until recovered or restored.