I think this is the crux of the problem, case law exists to help 'refine' and 'classify' a case ie. a crutch for an incomplete specification (the actual law). The 'workflow' is that the judgement on the current case informs the next ones and so on.

I'd prefer it if previous cases had a bearing on rewriting/amending laws, rather than influencing a current case, and if the feeling is that a current judgement was not correct, rewrite/refine the law taking the case into account.

That way a judgement does not require previous case research, this does mean that law needs to become a more well specified, and unambiguous language. We have decades of classification and computer language theory to draw from, and LLMs to stub-translate existing law, with exceptions going to a dead-letter-queue for individual appraisal.

Law should be understandable by the layman, else how does someone know if they have broken it?