What is the explanation for the archaic attributive particle が becoming a modern subject particle?

Question

While reading though Haruo Shirane’s Classical Japanese: A Grammar, I came across the following passage:

が started as an attributive case particle, became a subject particle, and then turned into a conjunctive particle. In the ancient period, the subject case particle did not exist. The typical structure was subject + predicate, as in 花咲く, or subject + adjective, as in 山高し. It was only beginning in the Muromachi period that the pattern 花が咲く and 山が高し, with が marking the subject, became standard.

While this makes sense looking at actual textual examples, I’m finding it difficult to understand the progression. Would anyone be able to explain what is the semantic basis for an attributive particle (or genitive case) taking on the role of subject particle (or nominative case)?

(As an aside, the same section in the same describes the uses of の as a case particle, which all correspond to the uses of が: subject marker, attributive marker, implied nominal, apposition. Is there some explanation as to why there would be two case particles which seem to serve largely the same purpose? And is this the underlying mechanism for the が/の alternation in modern Japanese attributive clauses?)

Answer 1

Early stages of Japanese did not have relative clauses, but it was possible to modify nouns with attributive verbs (using contemporary lexicon/morphology for ease):

咲く丘 a hill where something grows

I believe that from early stages, there was little restriction on the semantic role of 丘 in the action of 咲く, i.e. all this is really saying is "a hill that has something to do with some growing".

With no relative clauses, one would have to modify the whole noun phrase (not the verb) to say what is growing

花が[咲く丘] "a hill of flowers where growing is going on"

At some point, this was reanalyzed so 花が modifies 咲く, and we have something that looks like a real relative clause

[花が咲く]丘 "a hill where flowers are growing"

In the end, が is understand to be a general subject marker, not limited to relative clauses.

This is covered in the literature, as @snailboat mentions, but I don't have it in front of me, and I can't comment on the chronology without it. This is just in case you want some general intuition about the semantic progression.