I know the obvious answer is -- consensus.
It really is that, so you should think about that some more.
As I understand, they have the authority to change and broadcast the changes to the client software (the software that miners use).
Miners use all sorts of clients, and the authors of each independently push changes. With the advent of mining pools it certainly isn't as diverse as it could be, but neither is everyone using the same client.
If "they" decide to change the client's code in some "malicious" way -- its safe to assume that the majority of miners/customers will blindly accept it (at least till they discover the malicious intent -- which could take a few days, which may be too late!).
At this point, any push to the reference client's git repo has to have a discussion around it, and any obviously bad change would be called out within hours. That wouldn't even be an official release. It's very unlikely that significant people would upgrade to it.
More importantly, what exactly is their incentive here? Are they being paid for their services? If not, how "untouchable" (i.e., uncorruptible) are they -- do they need to be?
You'd have to ask each what their motivation is, but I'm guessing that many of them will say that they are just interested in the protocol and pushing its adoption. Some are paid by organisations that want bitcoin to succeed.
You haven't spoken of bugs. Bugs are far more likely. Every time there is a dangerous bug in popular software, e.g. the recent Apple SSL bug, there is accusations of a conspiracy. Simple incompetence is more likely, but it is impossible to know.
A developer could be enticed to insert a subtle bug into the default client in the middle of an unrelated change. This is a theoretical risk. A good way to mitigate it is to have a wide range of competing client implementations.
