If you are naive about data access, like in this case of inserting into an array without preallocation, C++ STL collections are pretty slow. I once wrote an algorithm that made heavy use of dictionaries in C++ and Python and the Python version was significantly faster. C++ just does not have fast general purpose collections in the standard library.
Sort of. I bet I could take a snippet of C++ and find flags which would slow runtime right the heck down myself.
Not that we needed it, but more evidence that the adjectives "faster" and "slower" are overloaded to the point of meaninglessness when discussing compilers and languages.