I guess the assumption is that the phenomenon starts at a point, spreads across the object sized l at c, and lasts for only a period of time t at any location. Therefore the pulse ends after l/c + t = T. By measuring T you can put an upper bound on l.

Example of this inference, from the Wikipedia on Fast Radio Bursts (but I have seen in many contexts):

> The sources are thought to be a few hundred kilometers or less in size, as the bursts last for only a few milliseconds

Does that make sense?