How can I silence `observer.long_lived_stale` false positives for session-scoped publications?

[welkinwong]

We received a critical performance alert:

Observer Leak: long lived stale
Observer on "users" has been active for 49.1h with only 0 updates/min.
This likely indicates a leaked subscription that was never stopped.

Metric: observer.long_lived_stale
Resource: users

We're on skysignal:agent@1.1.0, so the v1.0.20 auto-publish tagging (isAutoPublish, 72h vs 24h) should already be in effect. This observer fired at 49.1h, so the platform treated it as a named publication, not auto-publish.

In our app this is expected, not a leak:

• Logged-in sessions stay open for days (users leave a tab open).
• Several named, session-scoped publications observe Meteor.users for the current user / org members (for example an app-root publication that reactively follows organizationId).
• Those user documents often don't change for a long time, so the observer is long-lived with 0 updates/min.

We don't want to rewrite these publications as name: null just to get the 72h exemption, and we don't want to drop Live Query monitoring.

Question: what is the supported way to silence this false positive?

Specifically:

1. Can we exclude observer.long_lived_stale for a collection (users) and/or named publications, while keeping the other leak heuristics (growing observer counts, orphaned observers)?
2. Is there an agent or dashboard setting for this (allowlist, higher duration threshold, mute rule)? I couldn't find one in the agent config or docs.
3. If this can only be done on the platform side today, is a publication/collection allowlist (or treating connection-scoped named pubs like auto-publish) something you'd consider?

Thanks!