Method before suppression

Identity evidence in. Bounded labels out.

The worksheets reconcile only the anonymized rows you supply. They do not observe Android, query a phone, inspect Tasker configuration or infer notification text.

TimelineOrders callback rows by supplied sequence. A first strong tuple is new; a group-summary flag stays separate; a repeated tuple with equal post time is a possible replay; a changed post time is a possible update; missing comparison fields produce ambiguity.

BatchGroups rows on the exact observation time supplied, then separates unique child tuples, summary rows and repeated child tuples. It supplies no debounce window.

SnapshotComputes set membership for callback and active-snapshot strong identities. It never converts active-only rows into new events.

IdentityShows how package + ID may contain multiple supplied tag/user tuples. It does not claim a field set that is universally correct for every listener.

Current evidence boundary

Android documents grouped child notifications, group summaries, automatic grouping and notification identity fields. Tasker documents a New Only control, and AutoNotification documents an Ignore Group Summaries option. Those controls and fields are exact current supply; the retained job is the joined trace across lifecycle, batch cardinality, callback-versus-snapshot membership and compound-identity collision.

Problem evidence comes from public operator discussions observed on 2026-08-27. That evidence establishes problem fit only. Search volume, attainable ranking, external use, conversion, revenue and profit remain unknown.

What this cannot decide

It cannot prove an OS bug, plug-in bug, race condition or correct suppression rule. It cannot tell whether a notification should be ignored, whether two rows carry the same semantic message, or whether an automation action is safe. Review the actual listener documentation and test with reversible actions.