Hm, T indeed can be a reference type, so I think it is not possible to reword this paragraph separately

Hm. [basic.stc]/2: ... automatic storage durations are associated with objects introduced by declarations ([basic.def]) and implicitly created by the implementation (temporary objects).
[dcl.fct.def.coroutine]/13 says that the copies are the objects with automatic storage duration, but doesn't say that they are introduced by a declaration and I doubt that they are temporary objects. So [basic.stc]/2 misses this case.

Given that it is described how a coroutine's function-body is replaced, maybe we copy parameters syntactically, by injecting variable declarations initialized from parameters before promise-type promise promise-constructor-arguments ; and replacing id-expressions naming parameters with id-expressions naming the corresponding variables in the function-body? This way:

• there will be no need to fix [basic.stc]/2, because objects will be introduced by declarations
• it will be possible to remove this wording from [dcl.fct.def.coroutine]/13:

The initialization and destruction of each parameter copy occurs in the context of the called coroutine. Initializations of parameter copies are sequenced before the call to the coroutine promise constructor and indeterminately sequenced with respect to each other. The lifetime of parameter copies ends immediately after the lifetime of the coroutine promise object ends.

• it will be possible to remove «An identifier that names a coroutine parameter refers to the copy of the parameter.» from [expr.prim.id.unqual]/1

I think we should leave this alone for now, and wait for it to be handled by CWG's split of "object" and "variable".