This is incorrect, both particles have a superposition of k and ~k (to use your terminology). It is not the case that when the wavefunction collapses, we just suddenly find out the "internal state" of the particle. This "internal state" simply does not exist until the wavefunction collapse has occurred.
The "no faster than light" doctrine refers specifically to the transfer of information. And no information, in the strictest sense, has been transferred. Think of it like this: try to imagine how you could communicate with someone using quantum entanglement. If you can communicate something, then information has been transferred. But you can't. All you know is that if your particle is k, then theirs is ~k.
Now, you could make up a rule book that says "if my particle is k, then I'll do x, and you'll do y" but that rule book will have to have been shared between both parties before hand, at less than the speed of light.
Edit: should just mention I'm not an expert here, and articles like this are always interesting. Finding loopholes in our conventional understanding. But it's important to know just /what/ our conventional understanding is first to realise why things like this are important.