Thank you very much for your detailed exegesis, John. It is very informative and helpful.

Let me start at the back end. Maybe more later.

John Fairbairn wrote:
But if you do choose to speculate in depth I think two cautions apply: (1) establish why and for whom the text was originally written - it is hardly likely to have been for rules mavens; (2) do not impose modern mindsets on the ancient texts - in other words, do not make the facts fit the theory.
I do not wish to speculate in depth. I do wonder if there was an earlier form of go with the object of capturing stones. That is similar to pgwq's interpretation of the Dunhuang Classic, so I am interested in seeing what he has to say.

I do agree that it was not written for rules mavens or lawyers. More on that a bit later. And, OC, distorting facts to fit a theory is not good.
However, interpretation is impossible without theory. That's why there is an agreement that parsimony is desirable. I.e., it is good to have a parsimonious theory of interpretation. Occam's Razor was originally about keeping the number of entities posited by your theory to a minimum. That does not exactly apply to textual interpretation, but minimum description length (MDL) does. Now, MDL applies only given a descriptive language. So what descriptive language do you use? That's an open question, last time I looked.
I think that you and I both agree that context is important. One point where we seem to differ is that I consider the statement about Bent Four in the Corner to be important context, while you do not, because it is not in a section about rules. I don't that matters much, precisely because the Dunhuang Classic was not written for rules mavens. To say that Bent Four in the Corner is dead (or dies) at the end of the game tells us something about how the game was played. The 1949 Nihon Kiin rules included it in the rules, but other modern rules do not, conspicuously area scoring rules. Area rules do not need it.
For how theory affects interpretation, we need look no further than Chen's
The History of Go Rules (2011). In the second sentence he says, "we all know that territory scoring has logical flaws", That is a modern theory that has been thoroughly debunked. Not to defend the Nihon Kiin rules, but the territory rules of Berlekamp, Lasker-Maas, myself, Ikeda, and, I suppose, Shimada, as he was a mathematician, are all logically impeccable. There are a number of places where that theory seems to affect Chen's interpretations.
The Dunhuang Classic states: 碁有停道及兩溢者,子多為勝。OC, on its face it indicates that the player with more stones wins. Fair enough.

But how does Chen interpret 停道 (stop road). He states:
Chen Zuyuan wrote:
道 is an ancient Chinese go term. It is defined as an empty point surrounded by stones of the same color, and roughly corresponds to the Japanese "moku" 目. 停 in classical Chinese may mean an equal division or bilateral coordination. So "stop road" means that the empty territories of both Black and White are equal.
I'm sure he is right, that that is a possible interpretation, but it hardly seems definitive. And it fits with the idea that there is something wrong with counting territory. If the territories are equal, then you can ignore them. But to assure that they are equal, you have to count them.

Did stone counting in the modern era depend on equalizing territory? Surely not. This interpretation does not seem parsimonious, but it fits with the modern theory of the illogicality of territory scoring.
Chen continues:
Chen Zuyuan wrote:
Of course, we can imagine another simplification: if both players have equal stones, the one with more empty territory is the winner. Although the Dunhuang Classic does not mention that, C&IP {Carefree and Innocent Pastime Collection} does.
Now, this is a way of doing area scoring. It's one of the AGA methods of counting the area score. So far, so good.

But does C&IP actually mention that method?
Chen goes on to quote the C&IP:
Chen Zuyuan wrote:
C&IP wrote:
At the end of the game, Black and White need not fill up the board; the side with more empty territory is the winner.
It does not mention "equal stones" here, but this premise is certainly implied, because it can be taken for granted and therefore omitted.
There is a bit of a snag there. Can the play of equal stones be taken for granted? Chen examined the four relevant game records in the C&IP and found that, indeed, there was an expectation of an equal number of plays by each player. In addition, as with stone scoring, the scores do not fit the raw territory scores, but they do if there was a group tax. So far the reported scores do seem to be a method of calculating the score by stone scoring. However, play ended with dame left unfilled. For equal plays to guarantee the same result as stone scoring all the dame must be filled. Chen missed that implication.
Chen's conclusions include these:
Chen Zuyuan wrote:
2. Under territory scoring, deducting the remaining eye points {i.e., group tax} is not justified, so the rules of the Tang Dynasty were not territory scoring but territory counting, the counting method being stones scoring.
(Emphasis mine.)
This is a reasonable conclusion as far as it goes, as a group tax with territory scoring does not seem to be justified. Berlekamp's rules were only published in one book in 1994,
Mathematical Go, with an English and a Japanese edition. Berlekamp's territory rules naturally have a group tax. In fact, he went to some effort to eliminate the group tax in a variation of the rules.
Chen Zuyuan wrote:
3. From the term "each side" we can see that territory scoring must be based on "equal stones". But dame are not played, and it does not matter who makes the last move. The term "each side" is often not used, so it is easy to overlook.
An incredible oversight!

Chen saw that when the dame are not played out it does not matter who plays the last stone (move). But if it does not matter who plays the last stone, you don't have stone scoring. He failed to draw that conclusion.
Example:
- Click Here To Show Diagram Code
[go]$$B Dame unfilled
$$ -----------
$$ | . O X X . |
$$ | . O O X . |
$$ | . O . X . |
$$ | . O X X . |
$$ | . O O X . |
$$ -----------[/go]
Equal plays by each side, one dame unfilled, net territory score = 0.
- Click Here To Show Diagram Code
[go]$$B Dame filled
$$ -----------
$$ | . O X X . |
$$ | . O O X . |
$$ | O O X X . |
$$ | . O X X . |
$$ | . O O X . |
$$ -----------[/go]
Equal plays by each side, dame filled, net stone score = 1 point for Black. You can use territory counting to get the net stone scoring result, but only because the dame have been filled.
It is hard not to suspect that Chen had a mental block based upon his low regard for territory scoring.
Edit: Actually, Chen does notice that the result of Jia Xuan's Game is an 8 point win for Black, while the correct stone scoring result would be a 9 point win. His conclusion? Not that the players were actually playing by territory scoring, but that they were doing it wrong.
Chen Zuyuang wrote:
How can that be? A reasonable assumption is: If Black makes the last move, in order to have equal stones for each side, Black will remove his last stone. Once territory counting is adopted, dame are naturally not played. People tend to simplify habitually, so the last dame will be ignored.
(Emphasis mine.) Territory
counting, not territory scoring.