I've noticed that with single-question responses such as Amazon's "why are you returning this" drop-down, and it makes total sense there. A consistently random arrangement means that users picking just the top answer will have no more influence on the overall report than random noise. Makes less sense when available answers are the same for each question, such that the user might have mentally cached them to avoid rereading.
However, n=1. The timing might have been coincidental.
However, n=1. The timing might have been coincidental.