Plug In Plants
If you hate the idea of genetic engineering of plants, stop reading now.
Deep rooted perennial plants are better for the soil, sequester more carbon, and more efficient at extracting nutrients. Producing a new set of seeds/grains every year, while retaining the roots, means a plant can expend far more energy on the seeds and less on roots and stalks. Deep roots can extract nutrients further down, and exchange carbonate for phosphorus. Deep roots can establish symbiotic communities with beneficial fungi and bacteria. The genetic engineers are busy crossbreeding grains with perennial plants, tweaking on genes as needed to develop viable species. The task is difficult, with little success so far, but the value to the world is enormous.
But why stop there? For thousands of years, we have been grafting trees, especially heterozygous fruit trees like apples. Since apples do not breed true, the seeds of a nice sized, tasty apple will not grow into similar apples, but into wrong-sized, bitter, ugly fruit. So we grow root stock from those seeds, and graft cloned branches from tasty apple trees onto them. The vast majority of all apple trees branches are clones from just a few dozen apples.
Instead of laborious mechanical grafting, imagine fields of "plant ready" root stock, which self-grafts onto "top seeds" that are sewn annually. The root stock might be 20 years old, a well-developed investment; one year a farmer might grow grains on them, the next year strawberries, the next year pumpkins. Keeping the root stock healthy could be a challenge, but as new threats evolve, new amendment species could be added to the soil community to combat them.
Of course, these communities of plants will eventually die without active farmer input to add top seeds annually. Without this active process, the root stock will starve and uncomplicated wild plants will eventually take over. This behavior can be designed in, so that abandoned agricultural fields will revert to nature. The purpose of this system is not to spread the behavior into nature, but to protect wild nature from our genetic tinkering, to minimize the percentage of cropland, minimize runoff, and maximize the health of the soil used for crops.
In the long term, these complex agricultural systems will be designed to inter-operate with the inputs and outputs of cities (the most ecologically sound way to house humans), so we can close the nutrient and waste loops as much as possible with agricultural processes. This allows more attention and energy to be focused on the "leaks" - toxins in, nutrients lost - to complete the closure. If we reduce system leakage by a factor of two per decade, then effective amount of leakage for all time is two decades worth. That will allow nature to get about its business, without extra burden from humans.