Read the article, and they were careful. That being said:
a) The 8.3 km fiber optic, including Tx and Rx circuits, has some temperature coefficient of group delay. Since the GD is ROM 30 us, and they desire a couple ns cal error, then was the temperature at the cal times close enough to the temperature at pulse measurement times?
b) Would like to know more detail on how the digitizer time stamping was done. Concern is with front end latencies. My sense is they probably did fine here, but it would put everybody to sleep actually explaining it.
c) What if the proton pulse shape has a good-size temperature coefficient? Then will the pulse shape statistical treatment they did still work and not lead to errors? Because the pulse, 10 us long, might then give problems with the way they modeled it if it varies quite slowly.