In this vicinity I also noticed:

• The copy-constructor and assignment-operators for both flavors of vector give the rhs argument the name "x", which seems a bit silly; it could go unnamed if we wanted.

• vector<bool>'s move-constructor is not declared noexcept. This was deliberate; I don't really know why, though. N4258 made vector<T>'s move-ctor noexcept, and says briefly at the bottom: "No change in vector" (no rationale).

• The sequence containers and set declare constructors taking initializer_list<T>, whereas the maps and unordered_{multi,}set declare constructors taking initializer_list<value_type>. (This is almost entirely consistent with the hypothesis that for simplicity we use T wherever it's synonymous with value_type; except that unordered_{multi,}set breaks that rule.)