n8n 2.45 Turns On the Durable Scheduler. Check Your Intervals First.

A short field note for anyone running self-hosted n8n. Version 2.45, dated October 20, 2026 in the n8n changelog, turns the durable scheduler on by default for Schedule Trigger nodes. Before this release it was opt-in. If you upgrade without reading the notes, two things change under you.

What the durable scheduler does

The durable scheduler documentation describes it as running time-based workflows "from a database-backed queue" instead of from each instance's memory. Each upcoming run is written to the database as a row before it's due. A restart no longer drops pending runs. A run missed during downtime still fires if it's inside its grace period (default one minute, set by N8N_SCHEDULER_MISFIRE_GRACE), and after that the trigger's misfire policy decides: skip, run the most recent missed execution, or run the most recent per rule.

For most of the workflows we build this is good news. The in-memory scheduler's failure mode was quiet: a container restart at 02:00 and the 02:00 nightly sync simply didn't happen, with nothing in the execution list to say so. A database-backed run is at least a record you can see.

The change that will surprise you

From 2.45, "every N seconds" and "every N minutes" schedules run at a uniform interval counted from when the workflow was published. They no longer align to the clock. The docs' example: publish at seven seconds past the minute, and an "every 30 seconds" trigger fires at :37 and :07, not :00 and :30.

That matters if something downstream assumed clock alignment. We've seen workflows that run "every 15 minutes" and then compare against a report generated on the quarter hour, or that assume two workflows on the same interval fire together. After the upgrade they drift relative to each other by however many seconds apart they were published.

Cron-expression schedules are unaffected except in the edge cases the docs call out (leap years, 53rd weeks, daylight-saving transitions).

What to do before you upgrade

1. List your Schedule Trigger workflows. Mark the ones on seconds or minutes intervals.
2. For each, ask whether anything depends on the exact minute. If yes, switch the trigger to a cron expression (*/15 * * * * is clock-aligned and always has been).
3. Decide your misfire policy per workflow. A nightly CRM sync should probably run the most recent missed execution. A "ping Slack every 5 minutes" workflow should skip.
4. If you have N8N_SCHEDULER_ENABLED or N8N_SCHEDULER_TRIGGER_NODE_MODE set in your environment, both are now deprecated and n8n logs a warning at startup. N8N_SCHEDULER_TRIGGER_NODE_MODE=legacy keeps clock-aligned timing for now; the docs say it will be removed in a future version.
5. The "Skip Durable Scheduler" node setting is gone in 2.45. If you used it, the instance-wide N8N_SCHEDULER_ENABLED=false is the only remaining opt-out, and it's deprecated too.

The docs page doesn't mention n8n Cloud, so if you're on Cloud rather than self-hosting, assume the change arrives on their schedule and watch the release notes.

Our take

Durable scheduling is the right default. We'd have turned it on anyway for anything that touches a CRM. But a default that changes the firing time of existing workflows deserves a line in the upgrade plan, and this one is easy to miss in a release-notes entry headed by a feature name. If you'd rather someone else read those notes for you, that's roughly what our n8n automation retainers are.