What would you think of a smart contract that acts trustlessly 99% of the time, but has an escape hatch governed by a set of more or less decentralized/authorit

What would you think of a smart contract that acts trustlessly 99% of the time, but has an escape hatch governed by a set of more or less decentralized/authoritative entities in a multisig? We already have an escape hatch for generalized catastrophic events in the form of hard forks, but I expect in the future more fine-grained schemes to manually resolve local issues will become the norm.