I would suggest to do a sequential design process for every software development. Modifying update.conf is also in this group. You should try to make changes in DEV then really test it before asking for UAT and migrate it to PROD at the end.
Many people integrate version control to the policy. All hosts checkout the policy directly. You might consider to do that too if you like.
In my case, I don't change contents in update.conf much (once a year I guess). We freeze the code just for policy upgrade. Once I need to change, I do it in DEV, ensuring there is no thing wrong at all. As you may see, if there is a typo/human error, all your hosts might completely die, cannot update policy automatically.
I'm thinking to implement a double failsafe right now. One failsafe is to update policies regularly ran by cf-execd and another one for to only rescue failsafe if it fails.