I agree with the demand of testability for a scientific theory but I find Sabine’s formulation very vague, which is probably a good thing as making it more precise would probably make it more dogmatic and more one-sided. “Non-empirical facts” or qualitative considerations have always been a part of scientific method construed in the widest sense, and Einstein famously was indifferent to “verification by little effects”. Sabine argues effectively that observation is in fact present all the time, so why insist on a particular type of obsevation as an absolute necessity? Qualitative considerations could be sufficient observational fit for very abstract, very general theories that comply with non-empirical constraints as well. The argument for supplementary experimentation in the case of such theories is not made.
Surely the requirement of testability is not valid in itself, but rather as part of the more general requirement of realism, of actually describing the world. Some of Sabine’s formulations conflate too easily these two requirements, e.g. “The whole point of physics is to select axioms to construct a theory that describes observation”. A realist would say the whole point is to describe the world, and confrontation with observation is just one way to test this, not necessarily the most privileged in all circumstances. She also talks about describing nature, but it is not obvious that this is the same as describing observation. Some observation at least is erroneous or misleading, being based on false theories, and will have to be corrected or even jettisoned.
So I would reformulate this text’s concern as that of the increasing tendency to idealism in modern physics, and that this idealism underlies a certain indifference to testability. I don’t think Einstein’s indifference, insofar as it existed, was based on such idealism, but more on time considerations, i.e. that “good” verifications take time especially when the theory is very abstract. Relying on aesthetic criteria in this case would not be an absolute, but an interim measure necessitated by the fact that the necessary observations can sometimes lag behind theory.
The worry would then be that a properly interim measure could come to be taken as a permanent definition of science. This would lead to a new form of dogmatism, not so much of the content of a particular theory, but a methodological dogmatism where mathematical speculation reigns unchecked. On the other hand, no time limit to interim speculation can be specified in advance, as sometimes we don’t even know how to go about getting the necessary observations. Close fit to readily accessible observations can also be considered to be an interim measure, adapted to some phases of a science but not to all.
In conclusion, a new idealism utterly indifferent to testability would be a new obscurantism, and spell the death of science as we know it. But this is no reason to privilege one aspect of testability, experimental observation, and to erect it into an absolute criterion of scientificity. The requirement of realism must decide what is a relevant test or not.