Now, back on topic:
I agree with most of these, specifically:
1. POM keycaps.
2. Split keyboards.
3. Small spacebars, split spacebars, many keys for thumbs.
4. Ergonomic layouts.
5. Programmable Fn layers AND default layers.
5. Alps clicky switches.
Like jacobolus, I am doing something about it with the next version of my board design (split, ergonomic, angled thumb clusters, Matias switches option, POM keycaps option, programmable layouts with a GUI editor). I'm also experimenting with curved finger areas, capacitive switch ideas and buckling springs, but those may have to wait for prototype number 3.
IMHO, better to have both standard layout boards and genuinely ergonomic boards than to modify standard layout boards piecemeal (by splitting spacebars, or changing only character layout, or splitting anormal layout board into two pieces, etc). Means you can still type on a normal board when needed, but you get all the benefits of a properly designed board on your own PC. No mixing up of motor skills / memory, no need to "unlearn" anything.
IMHO it needs to be a "clean break", something so different that you don't use the same muscle memory for typing on the two. Of course, getting widespread adoption will be the biggest problem, no matter how much better a design is than standard. Look at Maltron. Lillian Malt even wrote scientific papers on the design proving its superiority and it still was only used by a small minority of people. I do think the design could be improved somewhat, such as being split and more compact / portable, more optimal use of the thumb clusters in terms of allowing combinations more easily, using Fn layers, using modern analysis on character layouts (not possible in her day, but fairly easy now), etc.
I believe both will need to coexist. Normal board designs will not go away just because something better is out there. People stick to the familiar, but sometimes will also try something novel. If people are assured that trying the novel idea won't affect their skills on a normal board they will be much more likely to try it. If it doesn't work for them, they've lost nothing. If it does they have gained. If the gain is enough, they will promote the concept to others. Eventually it may replace the normal design, but I don't see it happening by increments. "Normal" keyboard designs are too entrenched. Even a small change can be catastrophic to typing fluency and it's an irritation to adjust to. If you have to do this many times to get to a truly good design, many people will not make the effort, IMHO.
In many industries change happens by evolution, small changes that people can get used to as they're introduced one by one. I don't think this is the way to go with keyboard design, though. It needs to be a "revolution", not an "evolution".
