The point of the no-communication theorem is that if you transport one half of the entangled pair to Alice, and one half of the entangled pair to Bob, neither of them can use measurements on their half of the entangled pair to communicate. It's forbidden by QM as we know it.
If Alice can affect the initial state, of course she can use that to communicate something to Bob, because half of that state then gets transported to Bob. That's exactly as interesting as sending a qubit to Bob in the normal way, so: not very.
> one particle would change the initial state of the other wouldn't it
It wouldn't, or at least not in any way that is accessible to Alice or Bob. If Alice or Bob pokes their particle too hard it will stop being entangled, and there's no operations you can do on one half of the pair that will communicate any information to the other half. Widely-separated entanglement has already been demonstrated, the reason nobody has been able to use it to transmit information is because it's forbidden by QM and so almost nobody has tried, and nobody expects it to work because it that's not how entanglement is understood to work.
A lot of people have a mental model that entanglement means that there's an invisible see-saw between the two particles, so if you could just force one particle to be spin-down then the other one would be spin-up, and that would let you communicate. Unfortunately, the connection is much less durable than that; interacting enough with the particle to force it into a definite state is an observation, and observing the particle cannot transmit information (that's what the no-communication theorem says).
If it does somehow work, it will be because QM is totally wrong. Also it would immediately be used to cheat the stock market.