I am not as critical as Rob of the AB's Tuna 21.5 report, but I do have some concerns. Let me set them out briefly.
First, the Appellate Body said the following about the standard it applies under TBT Article 2.1:
7.153. We have already undertaken some analysis relevant to the first issue when we addressed the United States' appeal of the Panel's articulation of the legal standard to be applied in the second step of an assessment of whether a technical regulation accords less favourable treatment under Article 2.1 of the TBT Agreement. We noted that, in determining whether the detrimental impact on imports stems exclusively from a legitimate regulatory distinction, the "particular circumstances" of the case may inform the appropriate way in which to assess even-handedness in that specific case. While acknowledging the potential utility to the even-handedness assessment of an inquiry into whether the detrimental impact, and the explanations given for it, can be reconciled with, or are rationally related to, the policy pursued by the measure at issue, we explained that this is one possible examination that could be undertaken to determine whether a regulatory distinction is arbitrary or unjustifiable and thus not even-handed under Article 2.1. Yet, taking too narrow an approach to such an inquiry could lead a panel into error. To appreciate properly whether a measure is even-handed, an inquiry into the nexus between the regulatory distinctions found in the measure and the measure's policy objective may have to encompass also, or to be supplemented by, consideration of whether the differences in the treatment accorded to different groups of products by virtue of those distinctions, and the resulting detrimental impact, are disproportionate in the light of the objective pursued.
(See also para. 7.97) This is nothing new, because it comes from COOL, but the part about the "nexus between the regulatory distinctions found in the measure and the measure's policy objective," and whether the regulatory distinctions "are disproportionate in the light of the objective pursued," worries me. In the real world, with actual measures drafted by actual people, and influenced by actual special interest groups, I'm not sure how easy it will be to satisfy this test. As a result, once there is detrimental impact, it becomes very hard to defend measures as "stem[ming] exclusively from a legitimate regulatory distinction." I'm not saying that the "nexus" inquiry envisioned here is irrelevant; I just wouldn't rely on it as the sole basis in the "legitimate regulatory distinction" analysis. The inquiry should be a very broad look at the design, structure, architecture, etc. of the measure. Based on the findings in this case, I worry the broader perspective is getting lost.
Second, if you read through the crucial parts of the AB's reasoning leading to a finding of violation (at paras. 7.254-265), relating to the "determination provisions," the basis for the finding seems really quite small. It does appear that, perhaps, there are some minor elements of the U.S. statute and regulations that, to a limited extent, may not address harms to dolphins in some fisheries in the same way that harms in other fisheries are addressed. As a result, perhaps the U.S. measure is not perfectly calibrated to the risks to dolphins in different fisheries. But that lack of calibration is very subtle. And I'm just not sure how many measures can satisfy a standard that requires such a high degree of calibration.
And finally, while I'm complaining about things, let me offer a more general comment on the AB's Article 2.1 standard, which, to repeat, looks at "whether the detrimental impact on imports stems exclusively from a legitimate regulatory distinction." It has always seemed to me that this standard is a bit narrow and inflexible. I wonder if a reformulation that involves a more holistic assessment of these same elements, and clearer language, would be better. For example, you could look at whether (and to what degree) a measure has a detrimental impact on imports; and you could look at whether (and to what degree) the measure reflects a legitimate, non-protectionist policy. Then you would weigh and balance these two elements, to reach a final conclusion on whether less favorable treatment exists. I feel like this kind of analysis would be the one most likely to get you to the right result, and it would do so in a way that was more easily understood by, and thus had more credibility with, the general readership of these decisions. And it would only take a subtle shift from the current language and approach to get there, so it wouldn't be a radical departure from the existing jurisprudence.