The basic argument in favor of complexity is that it begets robustness, in effect redundancy, such that complex systems, whether physical or biological are more likely to survive.
I agree; point taken.
But wait, no less an eminence than brother Einstein counsels:
The fact is: complexity comes at a cost that is perhaps unaffordable. Complex systems are prone to chaotic behavior: relatively minor stimulus provokes multiple diverging responses. Is it possible to predict such stimulus-response: yes, physical systems always act according to their design. But, how do we know to do the predicting? There's no way to know, so largely it's not done, and the latent threat is always there, to cause a reaction at what cost? There's no way to know. Reducing complexity to the Einstein level is the only generic preventive strategy.