[schoen]

This seems clearly true for applications related to censorship resistance and certain properties of money and financial systems (if you explicitly don't want a government to make certain decisions in the systems), but maybe not for many other applications.

For example, is it likely that people don't trust Louis Vuitton to decide whether Louis Vuitton handbags are genuine? Isn't that the definition of a Louis Vuitton handbag being genuine?

In most cases where an institution is empowered to make a decision of some sort, the risk that they will later change their mind arbitrarily doesn't seem to loom very large. (I'd agree that it's great to have mechanisms for detecting this, although regular digital signatures and Merkle trees may suffice.)