Unfortunately there is a huge difference between a proof of concept prototype that shows the feasibility of an approach and a fully working version that could be released to the world.
The problem that I tackled in these research papers is after all a very old problem. It has been raised in the early '80s in the PhD thesis of Michael Plass and since then there hasn't been much in terms of theoretical further results, let alone practical solutions to it.
What I have been able to show in the recent work, is that the problem is solvable in practice (in a flexible/customizable way that makes it suitable for real life situations) and if doing so with a time penalty that is acceptable.
However, to turn that into a LaTeX package or kernel code (or even its own eco system) that can actually deal with all kind of real life issues one finds in normal TeX documents is something totally different.
E.g., it is a big difference to prove that something like footnotes are a minor nuisance that do not change the overall complexity of the algorithm and its time bounds to actually implementing footnotes into such an algorithm so that they work in practice, etc. for all the niggling issues that one encounters in real documents.
So bottom line is, I would expect several 1000 (yes 3 zeros multiplied with a factor strictly greater than 1) hours of work to make that happen and I simply don't have the time for that on an unpaid basis right now, because in some way I need to make a living (and that is not easier after I left HP). I was wondering at some point, if it would be a sensible project for crowd funding and perhaps I give it a try and see what happens.
Otherwise it will take a looong time simply because it is a huge problem (there is a reason that there are only a handful research papers between Michel's and my work).
Remark on @Marijn's comment above
I fear you are mistaken. Yes the actual problem as stated is due to the placement questions that are raised from physical restrictions of the medium (like book type pages). However, there is no way to "lift" them. You only replace them with other mediums to present your data. That medium might be more flexible and not have pagebreaks as problems, but then new questions arrise that are of a similar kind (and possibly more difficult to solve in fact). After all if your search space is too big optimization can be a challenge too. Take the simple problem of todays ebook novels (no math, nothing complicated, ...). They can do all kind of reflows but often end up with something that is really horrible quality-wise and better algorithms that do some level of optimization given current reflow text size etc would be a very hard problem.
Or take math: your medium whatever that is will have some sort of boundary (if only by what you can grasp with your eyes, even if the placement option to the left or right would be wider ... what does that mean to math formulas? Proofs written in a single 20 foot line just because you can? No, whatever the medium is and its possibilities you will end up with conventions that are needed to convey information successfully and these conventions are then posing restrictions (with or without flexibility) that you need to optimize against, either manually or through algorithms. But don't believe those algorithms being a single iota simpler.
In other words, if not you are distracted by the layout issues (whatever they are with your medium) your consumers will be, as it will hinder them.
cheers - frank