GUI TrayIcon crash using Windows task planner

That is pretty much the nature of any change; it hopefully fixes things but also introduces the risk of breaking stuff. Generally, I would say that the stable release fixes a large number of rare issues that causes backups to stop running, but I see the irony in having a “stable” that is less stable.

We take the stable release quite serious, so there is sufficient time to test and catch issues, but that also means that once a stable release is out, we generally do not update it unless there is a serious issue. Instead we start a new release cycle when the next set of goals has been reached. We have started the path to a 2.2 stable release now, with Release: 2.1.1.0 (Experimental) 2025-07-17.

That is the plan. We branch of at some point from the development, and then “mature” it through the experimental→beta→stable cycle. In this period, we can add in fixes that have been tested in canary, but the goal is to ship a stable that is unmodified from a beta release, so we are as sure as possible there are no unknown issues introduced for users who are not willing to risk issues.

The tradeoff here is that releasing a new stable has a major impact as all users (more or less) will be asked to update. If we update, say every 2 weeks, that gives update fatigue, and we have to spend a lot of time testing. At the current rate, we seem to be hitting longer than 6 months release cycle for the stable release, which was large caused by a the new UI needing to be adapted to all the quirks that were introduced over the years.

I think the need for stable releases dictates that we need to do it this way. If we are fine with merging in hot-fixes into a stable release, and then releasing it directly as a new stable without a beta period, we can do hot-fixing, but I think this is risky.

Release 2.3 is about make the install and update flows much smoother, including bringing back some form of auto update. Once that version is out, the situation will be different, as the work required by the end user is much less, and then we can consider a different release cycle.