From 23.0 through a year of point releases, the emphasis has not been new features but hardening the PJSIP stack, the WebRTC channel and media handling.
Asterisk 23 has now been through a full year of continuous maintenance since its release in October 2025. Read only the changelog headlines and the generation looks quiet — no architectural rewrite, no marquee feature. For anyone running production systems, that quiet is the best possible news.
From 23.0 to the current 23.x releases, changes cluster in three areas: edge-case handling in the PJSIP channel stack, negotiation compatibility on the WebRTC channel, and resource management along the media path. None of it belongs in a launch event, and every item corresponds to something that bites in the field.
Take outbound WebSocket connection handling, codec validation and passthrough options, or DTLS in RTCP-mux scenarios. Problems like these are hard to reproduce in a lab and only surface across complex carrier networks and endpoint combinations. Maintenance releases fill those holes one at a time, which is the same as buying every deployment out of a debugging session.
For a communications system, 'no surprises' is itself the most important quality metric.
Three lines are currently active. 23.x is the standard release where new features land first. 22 LTS is the mainstream production choice, fully maintained through October 2028. 20 LTS moved to security-only fixes in October 2026 and reaches end of life in October 2027.
The practical consequence: if you are still on 18 or 20, the upgrade plan cannot slip any further. The security window only covers high-severity vulnerabilities; functional defects and hardware compatibility issues will no longer be addressed at all.
Many teams adopt an LTS-only policy, which is sensible, at the cost of always learning about new capabilities a year late. The standard release is the test bed: new interfaces, new protocol support and new media handling all run there for a year before entering LTS.
A pragmatic approach is to track one standard line in a lab. You do not need it in production — you need it to answer 'did the features we care about actually mature'. By the time the next LTS arrives you have a year of observations.
Lifecycle management for an open-source communications system works like an operating system kernel: the risk of not upgrading accumulates, while the risk of upgrading can be controlled. Which you choose depends on whether you take regression testing seriously.