@tdpauw gave a nice talk about “Feature Branching is Evil” on the XP DAY in UKRAINE in November last year (the slides from a similar presentation a few days ago can be found here). The article “On DVCS, continuous integration, and feature branches” also lists some valid arguments against feature branching.

I have used “A successful Git branching model” in several software projects making both positive and negative experiences. The distinction between a master branch for releases and a develop branched used for integration helps you keeping your history clean. However, it works against the idea of being able to deliver at every point of time when applying CI/CD.

Combining your branching strategy with distributed code reviews (e.g. in combination with merge/pull requests) is another practice which helped me and my team members to share a common knowledge about our code base and keep a high software quality level.

However, I experienced more than once that feature branching can easy lead to have multiple long-lived branches which diverge and often end in a merge hell. That’s why I (and my others) recommend to work in small increments and integrate very often to ensure a lifetime of your branches.