I’ve only read the abstract, but it seems like an equally important question would be the time-to-fix for post-release bugs with code coverage vs without. My own experience is that bugs are easier to fix when the associated code has test coverage.
Intuitively: along with bug fixing, refactoring and behavior-altering rewrites are also facilitated by good test coverage. To my mind, preventing bugs is not necessarily the main purpose. Nonetheless, the results in this study are very interesting.