The risk vector isn't that the dev team will push "PR 666: add master key to all wallets". It's that someone will introduce a subtle bug (such as a side channel leaking your private key), which goes undetected for months or years, until it's revealed that they've been slowly siphoning off and selling bitcoin they weren't entitled to. In a scenario like that, people might not attribute value to the current chain or the months-old clean one.
You could fork the code without forking the chain. I'm actually surprised that there aren't multiple organizations writing software against the bitcoin data model. Transactions and blocks are pretty well understood at this point, and maybe having other devs writing test features and extensions could help push the community in a positive direction.
Anyone, anytime can fork the code, and thereby the block chain, and create a second version of the currency.
The question is: Which one do people attribute value to?