There are two different issues at play here. One is a replication crisis due to the original research being sloppy, using flawed assumptions, faking data, etc. This does happen and seems to be becoming more common. There are lots of predatory journals that will publish junk science and a lot of factors that are pushing researchers to ‘publish or perish’. The system is a bit broken.

There is a second issue that is more fundamental and more applicable to agriculture. That is a ‘replication crisis’ due to trying to apply reductionist science (i.e. I’m going to change one thing and see what happens) to a biological system in which that is inherently impossible. You can’t change one thing and not affect a bunch of other stuff. Ag researchers know this and we ignore it, otherwise ‘scientific research’ would be impossible. The bigger crisis in my mind is the fact that we try to use reductionist science to study living systems in the first place. Research should take a systems approach (i.e. I’m going to compare my system to your totally different system and see which is ‘better’). The problem is that research journals don’t like to publish that sort of thing and funders don’t like to fund it. Reviewers would nitpick it to death because there are too many things that aren’t controlled for and the comparisons aren’t ‘fair’. But almost all ag crop/livestock research is un-replicable. You can’t re-run the same experiment. There are just too many variables that change with time.