I learned from my quantum mechanics course that if you measure a quantum state twice, two things can happen:
1) You take the second measurement just after the first on. In this case, the result will be the same. The wave function has not yet "de-collapsed", so to say.
2) You wait a little before you take the second measurement. This time, the two measurements are not correlated, and the second result is again random.
Now relate this to the EPR-paradox. If you have two entangled particles far away from each other, and you do a measurement on one of the particles, you don't know if you got the result as a consequence of someone measuring the other particle or if it was simply random. This is the argument used, when someone tries to explain why no information is sent by entangled particles (and thus, they don't disagree with relativity or violate causality).
But what if those who measure the first particle were to continually do measurements? If you then try to do measurements (in intervals long enough that the second case described above should occur) on the second particle, you will keep getting the same result as a consequence of those measuring the first particle. Thus, it is possible to conclude whether or not those in the first lab are doing this kind of continuous measurements right now or not. Consequentially, information has been sent potentially with speed faster than light.