In addition to this, this wording does not cover the declaration of a reference being ambiguous with a the declaration of a function returning a reference.
Ex:

int x;
int& y(int(x)); 

I believe the intended interpretation is "the resolution is to consider the longest sequence of tokens that could possibly constitute a declarator to form a declarator". (In your example, the (int(x)) could be part of a declarator, so the declaration is disambiguated as a function declaration.)

I'd like CWG to look at this, in case my belief above is incorrect :)

My understanding is that the ambiguity arises because int(x) can be treated as a parameter declaration or an initializer, and the "resolution" is to consider it as a parameter declaration.

[dcl.dcl]/10 seems to suggest that anything that's not a typedef declaration or function declaration is a object declaration, so it does seem to cover declaration of a reference.

Both of your interpretations work, but I believe the wording should be changed in order to make it clearer as to what it means.

[dcl.dcl]/10 seems to suggest that anything that's not a typedef declaration or function declaration is a object declaration, so it does seem to cover declaration of a reference.

I think this is #1280