We're more likely to see one judge being out of line by being positive when everyone else was negative, if there was an error that that judge happened to miss. But it's possible a single judge saw (or thought they saw) an underrotation/edge error that the tech panel that did not call, and everyone else missed it.
This might be even more likely on elements like spins, where most judges might look away from the skater to the computer screen during the final windup, for the second or so it takes to input the GOE, and it's just on the exit that the skater has a slight mishap that could warrant reducing the GOE from positive to (at least in some views) negative final GOE.
Say someone did a 3Lz+3Lo. Very difficult. They under-rotate the +3Lo.
Someone goes "-1: it's a UR combo".
Someone else goes - "it's so hard, and the 3Lz was fully rotated. Also, everything else was fine. Plus, the combo is underrated in the current system. +3, because it's already getting scaled down by the penalty".
This is not at all to disagree with you, but rather to provide context. I'd fully support the second judge in this scenario, and frown at the first one, because IMO blindly following guidelines is never good.
According to the official guidelines,
These guidelines are tools to be used together with the minus GOE charts. The final GOE of a performed element is based on the combination of both positive and negative aspects. It is important that the final GOE of an element reflects the positive aspects, as well as any possible reductions that may apply. The final GOE of an element is calculated considering first the positive aspects of the element that result in a starting GOE for the evaluation. Following that a Judge reduces the GOE according to the guidelines of possible errors and the result is the final GOE of the element. To establish the starting GOE Judges must take into consideration the bullets for each element. If in an element there is any kind of mistake that requires a reduction, the starting GOE cannot be higher than +3.
This means that an element that was otherwise strong but that warrants a -1 or -2 reduction for one minor error could still end up with positive GOE.
So assuming all judges saw the same error, different judges might have different evaluations of how many positive bullet points to award (how high to start before reducing for the error) and how severely to penalize the error where the negative GOE guidelines give a range of penalties depending on severity of the error (e.g., -1 to -3 for "Euler executed as a stepover" or traveling in a spin).
For jump edge and rotation errors with tech panel calls that are shared with the judges, there are specific reductions judges are supposed to take AFTER awarding whatever positive bullet points the judge thinks the element deserves. Almost all the negative GOE reductions are ranges. So, e.g., for an element called 3Lz+3Lo<, judges are expected to subtract -2 to -3 from whatever they came up with for positive bullets (or non-called errors), and they can't start from higher than +3.
One judge could say "I was originally going to give this combination +3 because the rotation looked fine to me in real time, it looked effortless, there were steps leading into it, and it went with the music. But now I see the tech panel called the loop as under. OK, I'll takeoff -2, the less severe penalty required for that error. So I end up with +1."
On the other hand a different judge could have been less impressed with the preceding steps, effortlessness, and/or connection to the music and not have planned to reward any of them. And they might have seen the underrotation for themself in real time. So they might go as far as 0 as the starting point, -3 for the more severe underrotation penalty, final GOE of -3.
That could lead to GOEs ranging from +1 for the first judge to -3 for the second.
If the tech panel call had been q, then the GOE reduction range is -1 to -2, so the judge who was more impressed by other qualities could have ended up at +2. And a q call doesn't penalize the base value in any case.
If there's no call, the judge who didn't see any rotation doesn't have to reduce at all and can stick with their initial +3, while the judge who did see the jump as not fully rotated can take off -1 -- ending up at -1 if they weren't impressed enough by any other qualities to reward. So a range of -1 to +3 is possible if there's no underrotation call by the tech panel.
But if there is a call of any sort, a GOE of +3 is not justifiable according to the guidelines. The judge can't start higher than +3, and they are supposed to reduce by at least -3 for a << call, at least -2 for <, and at least -1 for q.
