I'm no expert but I think this is how it works. We'll call the particle pairs A, B, C, etc.
So when the researchers measure Local particle A (LA for short), the nature of entanglement means that Remote particle A (RA) must collapse into a specific known state. So they measure RA after that happens. The same is true for pairs LB and RB, LC and RC, and so on. Then they check statistically if the remote particles are all in the predicted states (should be 100% if this works flawlessly).
With enough repetition they can have very high confidence in the results. Of course those results must be communicated over traditional (non-entangled) channels that we already trust for reliability.
edit: typo