I suppose a lack of testing and an assumption that the action will fail so rarely that it’s not worth accounting for? But yes, such patterns make it hard to trust and efficiently use an interface.