Email Trigger (IMAP) fails silently on n8n Cloud — works only in manual execution (typeVersion 2.1, forceReconnect 5)

Same silent failure reported here, but in a very different environment: n8n Cloud v2.34.6, Skymail provider (Brazilian managed solution — not Gmail or self-hosted), IMAP trigger typeVersion 2.1, Resolved format, no custom email rules, and forceReconnect set to 5 minutes instead of the default 45. Low-volume mailbox, so the connection sits idle for hours between messages.

It worked for weeks and stopped triggering without warning — emails arrive and stay unread, the workflow remains marked as published, nothing appears in the executions list, the Error Workflow doesn’t trigger, and manual execution processes the exact same messages without any problem.

Two things I’d like to know:

  1. Has anyone correlated low forceReconnect values with higher failure incidence? At 5 minutes I run ~288 reconnections per day, and each one is a chance to fall into the handleReconnect path that silences both error channels.
  2. Has anyone checked if these crashes coincide with instance restarts from Cloud upgrades? A managed provider can hold the old connection slot for a few minutes after a dirty disconnect, so reconnecting right after a restart might be legitimately refused — and fall into that same silent path.

Hi @Carlos.souza Welcome!
Force Reconnect is what calls that branch. The node registers setInterval(handleReconnect, forceReconnect * 60000), so the silenced path is entered once per interval by design, not only when the socket drops on its own. At 5 minutes that is around 288 entries a day against 32 at the default 45, and each one is a full logout and re-login against Skymail, which managed providers throttle. Set Force Reconnect Every Minutes back to 45 or higher, that does not fix the bug but it cuts the exposure by roughly 9x.
On the restart question, the first connection does not go through handleReconnect. It runs at trigger setup, so a login refused right after a Cloud restart throws during activation and surfaces as an activation error rather than silence. To test the correlation, add a small workflow with the n8n Trigger node on Instance Started that appends a timestamp somewhere, then line those up against your dead windows.

That didn’t fix it. I switched the email to Gmail and now it works.