KH’s takeover mechanism addresses a different problem from moving listeners to a shared fallback. A local relay mount can retain its listening connections while an admitted source supplies the media instead of the relay feed. That changes operational control of the stream.

Enable only an intended takeover

The pinned parser recognises hijack inside a mount. This fragment belongs in the existing exact mount policy whose takeover is intended:

xml · Code example
<hijack>1</hijack>

It is not a complete file or a command to invoke on another server. Match the source’s path and authentication policy to that mount. An account allowed to replace an active feed has a stronger operational capability than an ordinary listener. Keep the permission narrow and verify rejection of an unintended source.

The changelog describes a source replacing relay delivery without moving listeners, followed by relay restart when the source leaves. It also records changes to takeover behaviour across KH revisions. Use the pinned implementation and a controlled test, not a general claim that all kh22-labelled builds switch identically.

Keep delayed shutdown separate

The pinned parser names its delayed source-exit control linger-for. Older changelog wording refers to a linger feature using different terminology. A lingering source can retain state while waiting for a replacement connection, changing when fallback or disconnect happens. Do not add a delay to “repair” unstable networking without understanding the effect on listener silence and recovery.

Exercise a complete handover

With a disposable relay mount, record its current origin and listeners. Connect an authorised compatible source, confirm the intended takeover and listen through the transition. Disconnect that source and observe relay recovery. Repeat with an incorrect account and with an incompatible test feed only in a safely isolated environment.

Check counters and metadata as well as the visible URL. A stable mount address does not prove continuous decoder-compatible audio. Preserve a known-good configuration with takeover disabled so the test can be reversed.