Struct ControlWatcher 

pub struct ControlWatcher;

Spawn-able helper that polls peek_interrupt and fires a tokio::sync::Notify once an interrupt message is queued. The orchestrator selects on the notify alongside the active worker run.

Implementations

impl ControlWatcher

pub async fn wait_for_interrupt( paths: MissionPaths, poll: Duration, notify: Arc<Notify>, )

Loop peek_interrupt every poll interval; when an interrupt is seen, fire notify.notify_one() once and return. Poll errors are logged and treated as “no interrupt yet” — the watcher never dies on a transient filesystem hiccup.

notify_one (never notify_waiters) is load-bearing: it stores a permit when nobody is waiting, so a fire while the run loop is between notified() registrations (or still inside backend.start()) is consumed by the NEXT waiter instead of being lost forever. Tokio also re-stores/passes on the permit when a woken Notified future is dropped unconsumed, so a select! race cannot swallow it either.