n8n 2.45 keeps scheduled runs across a restart

n8n's changelog dates this to 20 October 2026, in version 2.45. Schedule Trigger nodes use the durable scheduler by default, on every instance, with no environment variable to set. The scheduler stores each run in the database before it is due, so a restart no longer drops pending runs. A run missed during downtime still fires within its grace period. After that, the trigger's misfire policy decides. In a multi-main setup, one main claims each run.

Existing Schedule Trigger nodes switch over on upgrade, without republishing. Schedules of "every N seconds" and "every N minutes" now run a steady N apart, counted from when you published the workflow, instead of aligning to the clock. Hourly, daily, weekly, monthly, and custom cron schedules fire as before in everyday use. N8N_SCHEDULER_TRIGGER_NODE_MODE=legacy keeps clock-aligned timing on the durable scheduler. N8N_SCHEDULER_ENABLED=false returns the instance to the in-memory scheduler. n8n says both variables are deprecated. Poll triggers and system tasks stay on their current schedulers.

The same changelog page still lists stable as 2.43.8 and beta as 2.44.4. The GitHub release n8n@2.45.0 (opens in a new tab) is marked pre-release.

Who it's for: people running workflows on a schedule, especially a self-hosted instance that restarts.

Fits: Best Agent Builder. n8n stays the first pick for a team that already thinks in workflows. This is a reliability change on a pre-release, so the verdict is unchanged.