Plant growth & flavor perception may be scientifically hard to model at the mechanism level. As a heuristic alternative we can consider something like this:
Harvest 1 – Super tasting output of variety 1.
Harvest 2 – Avg tasting output of variety 2
Time course of nutrients for Harvest 1 was v_1,v2_2,….
Time course of nutrients for Harvest 2 was u_1,u_2,….
So we can try to move the u series in the direction of the v series and see what happens to variety2, and we can try to introduce the v series artificially with variety1 in an alternative environment. Other examples like that are possible. Either way, we need scientific guidance to say what is likely to matter and what needs to be controlled to give good chances. But we don’t necessarily need a futuristic level of biology & taste perception to achieve success.