Why can reserved vector::push_back() still be a performance bottleneck?
I understand that std::vector::push_back() is amortized O(1) because occasional reallocations are spread across many insertions. If I already know the approximate size and call reserve(1'000'000), repeated reallocations should…