The green revolution was largely the result of leveraging genetic improvement; using traditional breeding technology to improve yields^[1]. This came from exploiting naturally occurring variation in crops, identifying and testing outlier varieties to find new ones that perform better, and repeating this process to deliver continuous improvements in yield^[2].

The testing bit is key, and there is a pretty advanced network of breeding organisations led by CGIAR ^[3] who carry out trials of new crops across a range. While not exactly RCTs, crop trials are probably more advanced than RCTs as they involve many more repetitions, more environments, and more advanced quantitate methods to identify the crop varieties that work, and where. 

In short, the ag sector is very well set up to test and identify new crops that can deliver the next green revolution. so why hasn't it happened? Since the green revolution yields tend to increase year-on-year, so the next one will require a sustained increase in the rate of improvement.

What can we do to increase the rate of improvement in yields? It comes back to identifying and exploiting more variation in crops, and doing it faster (speeding up the cycle):

1. Better implementation of traditional breeding: although CGIAR etc. are pretty good at this, collecting environment-specific data for decision making in developing countries can be hard, and there are plenty of underrepresented species and regions.
2. Gene editing can be used to target specific genes. This eliminates the time and search effort in finding naturally occurring variation, and difficulty isolating that single trait through crossbreeding.
3. Speed breeding^[4]: combines intensive greenhouse farming with genomics (predicting performance based on DNA) to generate newer, better varieties, faster.

The second green revolution will not come from one breakthrough, but the cumulative impacts of many breakthroughs and steady improvements. 

Finally, to reiterate why breeding is particularly important for development and helping smallholder farmers: 

1. genetic gains compound: 1% better yield each year adds up quickly
2. new crops^[5] are extremely easy to implement compared to other tools or technologies (no new training, farmers just continue to do the same thing with new seeds)
3. new crops are cheap

1. ^{^}

   this was the beginning of year-on-year yield improvement, not a one-off change. . 

2. ^{^}

   but also other traits, for example ones that affect yield, like heat tolerance, or disease resistance - breeding goals can be extremely multipolar.

3. ^{^}

   some nice background: https://www.gatesnotes.com/How-CGIAR-is-feeding-our-future 

4. ^{^}

   speed breeding is not widespread, however, this type of approach is common in the earlier stages of modern breeding pipelines (before trials of promising candidates are scaled up)

5. ^{^}

   I've focused on crops but same is also true for animals - more resilient and higher yielding dairy cattle can be extremely beneficial for smallholder farmers. 

Copenhagen Consensus Center has a good paper on the cost-effectiveness of agricultural R&D investment, that might answer some of your questions.

One way might be to replace monoculture fields with more complex farming while everything gets managed by AI.