The premise, as before, is in part that as you learn, your brain not only changes structure but expands in those areas that are heavily focused on. It is also that the number of connections and their speed (different neurons work at different speeds) dictate how easily the brain can acquire new information and how fast the brain can process that information. (Unfortunately, denser neurons can carry more connections but are slower. However, this is what we see in computer networks, so it is no great surprise.)
On this basis, I am proposing that the optimal education is synonymous with the optimal brain, where I am defining the optimal brain as the one that confers the greatest freedom of choice in all life decisions, where freedom is governed by how easy it is to learn new skills, use existing skills personally or at work, and the upper limit to how far those skills can be developed. So, for example, how easy would it be for Fred Bloggs the Accountant to take up motor racing as a hobby or professionally? How quickly could someone pick up a new language and culture, if they wanted to move country? If the education you received hinders you in these tasks, it has limited your freedom.
Obviously, it's hard to teach lightning-fast reflexes. It's hard enough to teach someone how to drive a normal car on a normal road. I don't consider that something education hinders, unless it suppresses your movement to the point where your coordination skills and reaction times atrophy. That's not unheard-of, but it's unusual. However, most modern education underplays the value of physical education and that would most definitely qualify as hindering.
So we obviously want something well-rounded but intense. The intensity needs to not quite be on the level of the Japanese cramming schools, you need to learn and not just know. However, it needs to have a bit of oomph to it. People need to be running close to the upper revs limit of their brain but not red-lining it.
That's going to differ between kids, obviously, and between subjects.
I think that the best strategy would be to have five streams per class: -2 standard deviations, -1 standard deviations, normal, +1 standard deviations, +2 standard deviations. Kids should not necessarily be in their "natural" stream, they may work better when around people they know and like - the social factor - but they should not be pushed too hard or be in a situation where they're permanently bored.
That's not quite good enough. People shouldn't jump too far ahead or too far behind when switching streams. Let's say you have five teachers, each teaching the full spectrum of students, so you have five instances of each stream. You can therefore offset each of those five differently. Using the older UK norm of three terms of ten weeks (so 30 weeks), you could have each of the five classes offset by five weeks relative to where you'd expect to be with that rate of progress.
The gap between different speed streams will obviously vary, but it guarantees that there is never a gap greater than 2.5 weeks between those switching. That's not great, I regard it as being vitally important to maintain momentum, it's so hard to recover it, but it's the best you can do and still have a system that's remotely achievable. One of the reasons for having five bands is so that the gap will rarely get that big. One of the reasons for having five offsets is so that a kid with a-priori knowledge doesn't waste time and learns something new and interesting as quickly as possible. If the knowledge is in the middle of a term, they'd be able to directly leap onto new material because there's the same stream already going that has only just covered that information.
Being able to, though, doesn't necessarily mean they should. The whole idea of completely customizing the experience to each individual is that you have the freedom to do this sort of stuff. The whole idea of understanding the neurology and psychology of that individual is that you know when you should and when you shouldn't.
So, where does the neurology fit in? So far, this is just theorizing about how a process might work. Where do we get into how it DOES work?
Well, here would be a good point to do so. It should be possible to give a kid a high-resolution MRI (structural, not functional) scan twice a year. A 20% change in 4 years implies an average of a 2.5% change (which should be visible at 5T, certainly at 7T) every 6 months. Nobody knows what the safe dosage is for magnetic radiation, but nobody has died from any of the 9T scanners that are in use and there is so far no evidence magnetic radiation has any significant impact. Because any events that may arise from MRI are not entirely independent, the timeframe and manner of delivery does matter. The decay of the magnetic field shouldn't take six months, but you want a comfortable time for any intra-cellular repairs.
The higher the resolution of the MRI scan, the more clearly defined the change in structure. Because the brain is developing, it is going to change (unpredictably) anyway from all the other experiences, genetics, etc. The more clearly you can see what is going on, the better you can understand the mechanics. A standard medical MRI (2.5T) is far too insensitive. The images are blobs. You can't work with blobs. The cheapo doc-in-a-box MRI is 1.5T or less and is perfectly good for detecting brain damage but is about as useful for this kind of fine-scale work as popcorn is for deducing the shape of the universe. That's why you have to go to the upper limits. The 9T scanners would be great, but nobody is going to be willing to have a child put into a machine rated by medical agencies as too dangerous for medical work.
You will be never be able to 100% identify a specific effect to a specific cause, but you'll be able to see if the effect exists for there to be a cause. It would take a lot more than just one change set per student to establish the exact impact of learning, remember statistics is only valid for large numbers, and different tests require different numbers to be large. A lot of people screw up with ANOVA tests because they haven't got a random population, haven't tested for one, and haven't then checked to see if the correlation DOES indicate causation. You've got to have enough data for that.
Besides, your concern with cause-and-effect is limited to whether you can use the data to identify how the educational experience can be improved. This includes reducing any harm done by that experience. If the scans show that there is harmful under-development or over-development, or if there is any data available that might explain disturbing child psychology assessments, then the experience can be tweaked to counteract the negative.
(Child psychologists are important in all this, but as there is no current theory as to how an optimal kid would behave, they should not be too deeply involved. Psychology is not a particularly well-developed field anyway and has garnered rather a bad reputation due to cases of over-prescribing medications for economic gain and of other misuses of the profession. I'd rather keep them out altogether, but you need the data from them and teachers if you're to understand the feedback loop you're creating.)
The system should be monitored by evidence-based education specialists, so that as data is collected, they can establish what actually is relevant. My approach is mechanics-based, but not all mechanics are equal and not all mechanics are well-enough understood to know how they fit together (if they do). The neurology may not be significant over such short timeframes, the psychology may be irrelevant altogether, there may be too few streams or too many (although it's probably too few as there are a lot of people outside of +2 std dev).
I'll just sum up the rest, as it's all very basic stuff.
Kids should never be pushed too far beyond their attention span and the brain doesn't handle drudgery well. They should get plenty of breaks, as play is an important part of learning (something often forgotten).
There is strong evidence to support healthy meals (Jamie Oliver demonstrated that it reduces sickness in kids and increases attention span and intelligence, research into the Blue Zones have shown that it makes a huge difference to health). There is also strong evidence to support many smaller meals rather than one large meal (F1 driver Nikki Lauda detailed in his autobiography the importance of eating the right sort of meal for the job). These pieces of research are corroborated by research into the gut biome (which determines the health of the autoimmune system and affects brain performance) and the biochemistry of diet.