As your project grows, it's hard to forsee possible side effects of newly added features - and even fixed bugs can backfire in unexpected ways.
There is known solution to prevent regressions, and it's automated testing.
Regretfully, even if a project is good enough to have budget for automated testing, it's the first thing that is cut when something goes wrong (other issues take more time, new features are added without increasing the budget, other unexpected stuff).
On the bright side, there are "semi-ready" solutions that minimize the set up of automated tests (and give other benefits). I'd like to describe practical setup, based on several very different projects, hosted on different platforms, and it's based on Gitlab CI, Behat and Acquia BLT