Sounds great Sam. I think the tricky thing is defining what a regression means. For instance, if you are fixing a regression and introducing a less important regression then what? Master is technically undeployable in both states.
Real world example:
Also is it a regression if the tests do not pick it up?
It seems that if we pick up a regression 2 weeks later, it's not almost the less sensible/feasible to revert it but I'm not sure.
Just stuff to think about. I agree in principle we need to be more diligent and extreme when regressions happen.