The article title doesn't seem accurate? Although I'm sure they had a faster response time to help it doesn't sound like the person was saved as a result of the actions
For constant-surveillance companies like Facebook, a good way to justify the data collection would be to add automated systems to detect such problems. They already have algorithms in place for suicide prevention, it would be easy to increase coverage to stuff like seizure, heart attacks, and other problems that may correlate with long usage of computers. Even simple heuristics such as sudden keyboard depression and subsequent inquiry dialogs that are missed could help. Of course, ML on webcam vis-à-vis cats accidentally walking across keyboard would help to rule out false positives.