The way that we usually test these circuits is by looping them. That is, we activate an electronic loop in the circuit cards or CSU (Customer Service Unit - read: glorified modem) and we send patterns and compare what we send with what we recieve back. With a loop up, both should be identical.
The problem was on the end user's transmit side, so what I think was going on was that when the tester attempted to loop up the CSU in the Shark beforehand, they wouldn't have gotten any feedback, so their computers would tell them that the loop failed to activate. Meanwhile, because the recieve side was working fine, the CSU was in loop, but nobody ever took the loop down until my tester and I worked on it. This meant that the dial tones couldn't work because they were running off of the back end of the loop.