[ricardobayes]

I think the opposite is unworkable, namely just shipping new features and thus building on top of a murky foundation. I know a company that did this and basically needed to set up a taskforce and spend 2-3 months of the whole department's effort (around 50 people) on fixing a bug that that was low in the stack but they just kept building on top of it. So in my experience, a correct approach is indeed to try to eradicate all known issues in every release cycle by running and completing the regression tests. Yes if that means slowing down a feature release by 2-3 days then that's the cost of having a bug-free application. I read somewhere that users don't care about features if they encounter bugs, they abandon the product rather easily.