Sounds like a lot of codebases I've looked at.
Case 1: long stretches of "non-coding" DNA indeed are "useless", but then also a material and energetic drain.
Case 2: long stretches of "non-coding" DNA actually have a use, and are thus a proliferative gain.
Case 3: for some stretches case 1 holds and for others case 2 holds.
Suppose a specific stretch is questioned for utility: prepare a corpus of organisms with the stretch intact and with the stretch removed (so there is identical genetic diversity in both corpuses).
Then let a minority of "intact" organisms compete against a majority of "genome light" organisms, repeat a few times.
Also let a minority of "genome light" organisms compete against a majority of "intact" organisms.
If case 1 holds for a specific stretch: the modified "genome light" organism will have a selective advantage due to energy and materials savings when duplicating genomes.
If case 2 holds for the same stretch: the unmodified "intact" organisms will have a selective advantage.
We will likely continue to discover ways in which non-coding DNA is used by life, however there is no question that non-coding DNA is far from "useless" and hasn't been for some time.
Within non-coding DNA there do exist some sections with no known biological function which some people call "Junk DNA" however, there is much disagreement about this, and we have only relatively recently begun to directly image structures on the scale of DNA and proteins in situ via cryo-electron-microscopy, allowing us to study the mechanisms and motions of biological machinery frozen in action. DNA and cellular machinery is still far too complex to simulate fully, so CEM is one of the best available tools for studying it. For those reasons, and the fact that the percentage of what folks refer to as "junk dna" has steadily dwindled over the years due to discovery of these functions, it's reasonable to expect we'll discover more.
Hmm, I'm pretty sure that's not how upper bounds work.
Seems like the purpose is to keep reader confused about some point to maximize time spent on page. And I'm quite certain LLM can do a lot better