This project has steadily made two releases per year since this incarnation began. After each release, we receive numerous bug reports within the first month or two. While the time-to-response for fixes on these issues has generally been great, users are made to wait 4-5 months before a fixed version is available (if they're not willing of if it's inconvenient to build from Mercurial). By then, more work has been done that has likely caused other bugs. Perhaps some of the recent complaints about quality are rooted here.
In a larger project, this might be solved by betas, release candidates, bug bars, escrow, etc -- but we don't really have time for any of that. Instead, I propose bugfix-only releases targeted 1-2 months after each initial feature release, while features and more complicated bugs get started in a new branch (assuming work begins before a bugfix release is made). After two months, reports tend to slow down anyway unless a distribution has picked up a new version. This isn't a major change from the current process as these bugs are usually addressed fairly quickly anyway.
Finally, most JOE users obtain it through their OS package repositories. If possible, we should encourage package maintainers to pick up these bugfix releases -- or at least number them X.Y.1 because distributions are more likely to take a new revision rather than a new minor version in the lifetime of an OS version. It might be wise to use the timing of major distro releases to schedule our timelines as well.
Perhaps Mr. Super Packager joy/shallot could provide some perspective on the last point.
I like the idea of one bug fix release per month as long as there are changes. We could have a release branch for bug fixes and keep default for the latest mostly working features (after they are ready in a feature branch).
In the past I had received pushback on having too many releases- the claim was that this annoyed the package maintainers. But maybe bumping Z in X.Y.Z would be more acceptable, I don't know.
I kind of want to release now, to pick up the segfault when editing shell-scripts fix.
Your branching strategy makes more sense. And yeah, the sh.jsf fix was on my mind here. I also broke a couple of things in the windows release (doesn't pick up user colors files, write access detection broke). Things like that.