I don't think this is totally unreasonable, nor unique to judges.
For example, the developed world rolled out school lunch programs as a way to improve academic performance, which at the time of implementation was controversial.
Skipping lunch is bad for school performance, but judges aren't skipping lunch. Judges eat their lunch at a regular scheduled time and so they can naturally adjust their eating habits to get them through the day without experiencing discomforting hunger.
that's exactly what the original study was proposing as an effect:
> They found that the probability of a favorable decision drops from about 65% to almost 0% from the first ruling to the last ruling within each session and that the rate of favorable rulings returns to 65% in a session following a food break.
it's not unreasonable as an original hypothesis; and it's good that we're testing it and finding out later that it's wrong. but the base hypothesis is not particularly egregious.
For example, the developed world rolled out school lunch programs as a way to improve academic performance, which at the time of implementation was controversial.