Proposed resolution:

This wording is relative to N4659.

1. Modify [utility.syn], header <utility> synopsis, as indicated:

   […]
   
   // [pair.astuple], tuple-like access to pair
   template <class T> class tuple_size;
   template <size_t I, class T> class tuple_element;
   template <class T1, class T2> struct tuple_size<pair<T1, T2>>;
   template <size_t I, class T1, class T2> struct tuple_element<I, pair<T1, T2>>;
   template <class T1, class T2> struct tuple_element<0, pair<T1, T2>>;
   template <class T1, class T2> struct tuple_element<1, pair<T1, T2>>;
   
   […]
   

2. Modify [pair.astuple] as indicated:

   tuple_element<0, pair<T1, T2>>::type
   

   -1- Value: The type T1.

   tuple_element<1, pair<T1, T2>>::type
   

   -2- Value: The type T2.

   tuple_element<I, pair<T1, T2>>::type
   

   -?- Requires: I < 2. The program is ill-formed if I is out of bounds.

   -?- Value: The type T1 if I == 0, otherwise the type T2.

3. Modify [variant.helper] as indicated:

   variant_alternative<I, variant<Types...>>::type
   

   -4- Requires: I < sizeof...(Types). The program is ill-formed if I is out of bounds.

   -5- Value: The type T_I.

[ 2017-06-12, Moved to Tentatively Ready after 5 positive votes on c++std-lib ]

Instantiating tuple_element with an out-of-bounds index requires a diagnostic for tuple and array, but not for pair. The specification requires an out-of-bounds index for pair to go to the base template instead, which is undefined.

Similarly, instantiating variant_alternative with an out-of-bounds index violates a requirement, but it is not required to be diagnosed.