Then you could as well argue peer review should not look at methodology, but only for fraud, plagiarism and ethics violations. A retraction can in some cases be seem as a delayed failed (broader) peer review.
Methodology is part of fraud and maybe ethics. It's also required in the future for meta analysis. It might be that published papers disagree on the topic. If there are problems with methodology you can't really use the data as it's noise (at best).