You follow the process to have the decision reviewed and you make your case. Even if they reject what you propose you're still able to try to convince others that your idea is better and, assuming that you have a democracy, you can have the rules changed.
> If they're allowed to stop you then you can't demonstrate that their policy is in error
I struggle to imagine a scenario where there's no possible way to determine if an error was made or if a change might be an improvement without just doing it anyway. If it isn't plainly obvious that your idea would be better and there are also no facts, research, examples, models, etc. that give support to your idea you should expect to have trouble convincing others of how great it is. When something only impacts you personally, you can blindly try things without any reason to think it might be better all you like, but people who are making decisions that impact large numbers of people, whole communities even, should require some kind of evidence before changing things for everyone. Ideally, there'd be a process in place so new solutions can be proposed, promising ideas can be explored, researched, trialed, tested, then implemented.
> Doing that sort of thing suppresses diversity and breeds monoculture, which is how you get catastrophic systemic failures.
Sometimes, but not always. What you'd call monoculture can be really great. It's why international standards and internet protocols exist and are followed. It promotes interoperability. It'd be a lot easier to write an app to help people keep tabs on their power usage and charges if there was just one company putting out bills and an API. Even with a single company providing a service they can still make good use of diversification to limit their risks. Companies with tons of competitors often take those kinds of precautions to avoid single points of failure within their operations. The last ISP I worked for had over 15 types of modems in the field from a variety of manufacturers. Some were being trialed, some were being phased out, but a firmware problem or failing component impacting one model wouldn't take down every customer. Companies tend to try to standardize where they can though.
There are certainly some trade offs involved with having one organization handling things, and it's not ideal for every type of service, but things like utilities are good candidates because of the scales involved and because the "product" being delivered is the same. When power companies are doing what's expected of them nobody has a preference for the electricity that Company A provides over what Company B delivers. I wouldn't want a single company making all the pizza though, because in that case variety matters.