This is my opinion, so take with a grain of salt:

I believe it's entirely possible that the scientific method breaks down past a certain level of system complexity defined somehow by thermodynamics.

This would in part be due to the infeasibility of running the proper experiments to understand the effects of single variables when tens of thousands of variables might be changing at the micro level.