So 1/4 times it does not introduce a regression. That's still pretty bad imo. If 1/4 commits introduced regressions, what would your team do?

We are talking about regressions, what once worked no longer does, and should be measured in 9s