I liked how, back in the day, when the figures for the papers were in .eps format, I could just change the line-width on the plots by just editing the file. Now they are mostly pdf which is better for the file size, but worse for tinkering, or removing bits from the plots.
I worked in a mail house and really liked the editable nature of PS files that we got. We could run them through data processing and easily add/remove/adjust data. Sometimes the clients would ask us to do it in our monthly work because engaging their internal IT was too much effort, so we simple edited the files in the same way every month automatically before printing.
With the shift to PDFs though, it's way harder. Your simplest course to 'remove' data for printing is often drawing a white box on top of it but that but it won't reflow anything. It's more of an append-only format, core editing is possible but annoying, you need to use an actual PDF reader tool for text extraction and parsing rather than a Perl regex, you can't just reorder pages by cutting/pasting around the DSC %%Page: comments, etc.
I've just tried out of the interest making a plot in matplotlib, (1) save it to eps and (2) save it to pdf then convert to ps. And the original .eps is pretty readable, but the ps converted from pdf has some kind of "stream" encoding string. I.e. it's not really editable.
I think you're describing stream compression, which appears to be utterly ubiquitous to the point that the knobs to turn it off have either technically bitrotted or faded well below our collective awareness threshold.
Alternatively you might use something like `qpdf --stream-data=uncompress input.pdf output.pdf`.
qpdf (http://qpdf.sourceforge.net/) seems to be a good recommendation to add to the "mentally index the entire man page to know what it can do" list for anyone interested in PDF in general.
