Yes, it's possible to flip the spin of a particle. But you can't determine the result of the initial observation - it will be up or down with whatever probability the initial state would give - and once you have made that initial measurement the entanglement is broken, so anything you do to the spin afterwards will have no effect on the distant particle.
That's the problem: the distant observer cannot even know whether you have made a measurement of your particle, nor what direction you measured in. You could agree in advance when you would do this and what axis you would use, but that would only allow you to infer what the other person had seen (if you use the same axis, not even that if not). As far as communication goes that's like trying to send a message by flipping a coin and telling the other person which way it lands: since you can't control it it can't encode a message.