I agree that the human patient case studies are not a strong example of science, but think they should count at least as weak examples. The hypothesis being tested would be that certain solutions can be successfully introduced via perfusion (various methods tested) to minimize morphological and/or biological damage to the brain, and under what circumstances. The results can be measured e.g. afterwards by CT scans, measuring concentration of solutes in venous return, or microphone acoustic measurements of ice formation / cracking.

Arguably, the rabbit kidney studies of Fahy, Wowk, et al are motivated primarily by cryonics and only secondarily by the desire to transport lab-grown organs more effectively. (On the other hand, lab-grown organs are more promising candidates for this than donor organs, since they can be genetically and physically modified.)

One challenge to proposing animal research is that it could raise animal rights concerns, which could be exacerbated by cryonics' delicate political position. One way around this would be to focus (for now) on testing models that do not involve reawakening the animal. For example, Peter Gouras recently proposed on Cryonet that the mammalian eye could serve as a good non-revival model. The animal would be euthanized, and the eye would be subsequently cannulated, perfused, and cooled. When re-warmed, measurements of the electrical effects of a light flash on the retina would be used to validate its functionality.

Science most definitely does not mean: "Let's try something random today and see what happens".

That does seem to be how mathematics works, though.

They did some research on non-human animals in the past, but AFAIK, they didn't publish much.

You seem to concede that you were wrong and they have done some research, even if most of what they do isn't. That was my sole objection, so I think we're on the same page now >.>