This is the failure the experiment caught in the act. One innocent sentence, and the model read it three ways.
Ambiguity
A sentence that can honestly be read more than one way. A model rarely stops to ask which you meant. It picks a reading and writes working code for it, so nothing looks wrong.
Between people, ambiguity comes back as a question. From an agent it comes back as behaviour. "Updates an order's status" gave us three kinds of program. All of them type-check, and they agree on the obvious cases, so nothing tells you a choice was made.
Say it to your agent
“Before you start, quote back any sentence in this brief that could be read two ways, and give me both readings with an input where they differ. Do not pick one yourself.”