How scheduling works

Ten-minute passes, one shared queue across Macs, leases, and the three pause scopes.

How scheduling works

Posting on autopilot means a pass that runs without you, a queue every Mac can share, and pauses that mean what they say. This page explains the arrangement; the words and commands are in Job options and schedules.

One pass runs every ten minutes. It wakes, claims the rows that are due, runs them one at a time, writes the ledger, and sleeps again. A pass that finds nothing due costs nothing and changes nothing. The service owns the passes while the app is closed; your terminal owns the pass you trigger by hand, which never waits for the interval.

Several Macs share one queue through leases, not locks. A claimed row says running with a lease timestamp, and the holder renews the lease as it works. A row whose lease lapsed is some dead machine's, and the next pass reclaims it. That is why a row's Status never reads as a lock, and why two Macs never work the same row: the claim statement reads the lease alone, atomically, and only one claimant wins it.

Three pause scopes cover the three reasons to stop. This machine's file stops this Mac and nothing else, for maintenance on one machine. The organization-wide pause stops every Mac sharing the store, with a reason every banner repeats. One row's pause holds that row everywhere until resumed. A row is claimed only when none of the three says paused, and quitting the app touches none of them: pausing is always explicit, resuming always named.

A row that did not run is never a mystery. schedule why reads the claim the pass would make: paused at some scope, not yet due, lease held by a living run, or eligible and waiting for the next pass. The popover says the same thing in fewer words: due now, running, paused, or needs you.

Last modified October 9, 2026: docs: sync from master (5f4e718)