Alerts re-notify on every evaluation while firing

A firing alert sends a notification on every evaluation interval for as long as the condition persists (e.g. every 5m).
For a sustained condition like high node memory, this produces a continuous stream of duplicate notifications until it resolves.
There's no setting to notify only on the OK→ALERT transition, nor a re-notification interval / cooldown.
This option would reduce massively the amount of alert we are receiving at the moment.

Current behavior (packages/api/src/tasks/checkAlerts/index.ts, HyperDX 2.27.0)
The ALERT notification is sent unconditionally whenever the threshold is exceeded:

if (doesExceedThreshold(alert, value)) {
  history.state = AlertState.ALERT;
  history.counts += 1;
  await trySendNotification({ state: AlertState.ALERT, ... });
}

By contrast, the resolve notification is edge-gated, firing only on the transition:

if (previousHistory?.state === AlertState.ALERT && currentHistory.state === AlertState.OK) {
  await trySendNotification({ state: AlertState.OK, ... });
}

Version: HyperDX 2.27.0

Start in packages/api/src/tasks/checkAlerts/index.ts, where ALERT notifications are sent on every evaluation and OK notifications are transition-gated. Trace the alert history and notification flow to define the OK→ALERT behavior and re-notification or cooldown option; done means sustained alerts no longer produce duplicate notifications on every interval while resolution still notifies.