I'm pretty stubborn about preferring the 'unstable HEAD' methodology as used here, which I have used successfully over many repo types (back to RCS over NFS over international WAN, eek!) and over several decades, and with multinational teams. Branch management is hard.
We're finally cranking up the CI which means that the master should rarely be broken seriously or for long or in really bizarre ways. It's been a long time coming and is not by any means done yet.
But I'm veryt happy to consider, alongside our production branches, some 'stable dev' branches for those feeling less adventurous! I can see some definite virtues in that, and they can be anchored in, for example, library releases.