Imagine you wish to post-process a given field on a thousand of pressure levels : the software will fail because the maximum number of output levels is limited to a reasonable value.
However you can easily overcome this limitation by slicing the list of post-processing levels : that way you would submit a bunch a jobs, targetting the same output file. Since the output file is not sequential but indexed-sequential, the file will not be overwritten at the beginning of each job : instead the fields will be appended to one another.
You can also use this trick to append fields to your own input file : to do that you just have to copy your input file to the output file before starting the post-processing job.