Open code and data. It's de rigueur in a lot of the more prominent social science journals for obvious reasons.
I will say, however, that we need to be very clear about making a distinction between bad methods and intentional malfeasance. Academic research generally involved a small team of researchers, at best. The code and amount of overall effort involved for what the earlier poster described is extremely intense. For better or worse, code QC simply doesn't exist in most departments for largely financial reasons.
You can generally expect that - most - efforts fall short somewhere, simply because the odds of not making a mistake or two in a 15K+ code base with one or two non-professional programmers are almost nil.
The hope is that another team in the future will tackle a similar task and, in the process, obtain results that are or are not consistent with what were previously published. Thankfully, if the question is interesting or impactful enough, and climate research certainly is, the aggregate body of evidence will eventually force out unintentionally bad research.
Of course, open data and code requirements facilitate this.