A particular file was chosen as extreme edge case. Huge PDFs are common, but their size is mostly because of images. Here it's "pure vector graphics", and I experimented with how soon CAM::PDF will fail because of "out of memory" when building "page content tree" (as opposed to "just open and parse objects and report this or that").

I.e. with this edge case I'm checking if it's practical to "build tree" considering speed and memory requirements, and if some improvements (still just ideas) can make it use less memory. And with "R"-less stream I couldn't check that because it would take hours and I didn't understand what's going on. I mean, to test modified parser, maybe ordinary distributed test files are OK.