Connection is not torn down at the end of a calling workflow

[pilif]

This is not a problem when using GitHub-hosted runners which will be destroyed at the end of the job, but in case of self-hosted runner, a runner will stay running after a job has completed.

As this action does nothing to kill the tailscale daemon nor manually disconnect, I would say that in such cases, the machine, even though intended to be connected only during the runtime ob the job, stays connected to the tailnet.

I don't think this is currently possible using composite actions (https://github.com/actions/runner/issues/1478), so you might have to migrate to a TS/JS based action or at least put up a warning in the README that this action is not to be used with self-hosted runners, at least not without some further manual intervention of the user.