question about a bison documentation example of resolving shift/reduce conflict

Question

When I read the bison document, I found a example:

%token ID
%%
s: a ID
a: expr
expr: %empty | expr ID ’,’

and the document tells us it has a shift/reduce conflict:

First example: expr • ID ’,’ ID $end
Shift derivation
$accept
→ 0: s                             $end
    → 1: a                     ID
        → 2: expr
            → 4: expr • ID ’,’
Second example: expr • ID $end
Reduce derivation
$accept
→ 0: s                     $end
    → 1: a             ID
        → 2: expr •

we can found this conflict is due to when stack is expr ID ’,’ and the next symbol is ID, it can reduce to expr (because s: a ID), or shift ID.

then it gives a solution:

expr: ID | ’,’ expr ID

Attach the original document sentence:
This conflict is caused by the parser not having enough information to know the difference between these two examples. The parser would need an additional lookahead token to know whether or not a comma follows the ID after expr. These types of conflicts tend to be more difficult to fix, and usually need a rework of the grammar. In this case, it can be fixed by changing around the recursion: expr: ID | ’,’ expr ID.

BUT why ID | ’,’ expr ID ? This directly modifies the terminal that can be matched on! Just eg. ID , ID , ID it cannot work!

I wonder why the documentation says, find a grammar that resolves the conflict without changing the match.

ps. this is the document link and it's ver. is 10 September 2021, Bison Version 3.8.1, p146-147