The problem here that people are failing repeatedly to grasp is that science does not actually have an overarching "THEY" who will police the system and prevent rubbish papers being published. Science works on peer review, meaning that the peers (i.e. other scientists in the same field) of the person submitting the paper get to police what is written in the paper. This works if the peers are more or less skeptical, and fails badly if most peers either subscribe to or have a vested interest in supporting a set world-view. In the case of climatologists, a prophesy of doom is just what is needed to keep the grant money rolling in, hence there is a presumption that any paper that agrees with this orthodoxy must be correct.
The second bug in the system is how research is funded. Research grants typically fund a post-doctoral researcher (a person who has taken a degree and a doctorate in the subject) to work on the topic for three years. Typically this means a year to work out how to do the work, a year to actually achieve something and a year of blowing one's own trumpet to try to secure another post-doc posting. This is, as you might imagine, an inefficient way to fund research.
The third bug in the system is a lack of research support for climatologists. To be a good climatologist you have to know a lot of physics, a good deal of mathematics and a smattering of biology, meteorology and so on to actually have a vague clue how a climate system works. To model climate you need a doctoral-grade level of skill in quite sophisticated programming, including how best to use massively parallel machines. If instead of looking at the emails from the University of East Anglia leak you go and look at the code, and the HARRY_README file, then you see how this is working or rather not working in practice. The UoEA code was mostly cobol. Mostly, because Cobol sucks rather for manipulating strings; the coder used shell calls to handle these bits. You or I would likely use Perl or Python to do the same thing and would make a much better job of it, but the coder in the UoEA case was self-taught. Two words that ought to send a shudder through anyone tasked with debugging and maintaining code: Self. Taught.
That code was a mess. A complete and utter dog's dinner. The original coder was learning by making mistakes, the sorts of mistakes that get drummed out of newbie coders in their first year of a degree. Things like a sub-program failing silently, instead of screaming blue murder on STDERR, for instance. Things like not keeping adequate backups. The UoEA lost their input raw data, because they didn't have enough storage. Once the raw data was gone, they had no way to start again from scratch.
Climatology is like particle physics was fifty years ago. Particle physicists started off as jacks of all trades, and very quickly realised that an army of engineers and computer techies was needed to support a few researchers. Climatologists have yet to quite realise that they need a small army of coders and computer sysadmins to provide them with the kit and the code to run their simulations, and no, this support cannot be done by the hired help on a shoestring. Until they realise this, we will not be able to trust their results.