It might be weird, but it might be the result of forces in equilibrium:

e.g. one explanation might be that time spent prompting is time spent not improving, or learning, so it would not be surprising then that the productivity gains might be similar in scale to the productivity lost from learning / improvement that never happens.

another explanation might be that prompting just delays work that will have to be done later when fixing errors. When encountering an error that the LLM doesn't fix on the first attempt the human-in-the-loop doesn't understand the problem very well: they will either pull the slot machine for a while until they hit lucky, or grit their teeth and sit down to understand what the code is doing in the excruciating and boring detail required to find and solve bugs, which is exactly the work that they had offloaded onto the LLM earlier.

Sure, it’s possible. But I don’t see anything that could constrain those things to be of similar magnitude. If they just happen to cancel each other out by luck, that would be pretty strange.