A workflow we looked at recently reads new leads out of Gmail, logs them to a Sheet, and fires a Slack ping. It ran clean for six days. On day seven it just stopped — no red node, no alert, the trigger simply quit polling. A one-line reconnect fixed it, and it ran another six days before dying the exact same way. That rhythm is the whole diagnosis.
If your n8n-to-Google credential breaks on a cycle that smells suspiciously like a week, the problem isn't your workflow logic. It's the publishing status of the OAuth app n8n is using to talk to Google, and the fix takes about two minutes once you know where to look.
The cause, stated plainly
Any OAuth app still sitting in "Testing" status in Google Cloud Console gets refresh tokens that expire after seven days — full stop. Google documents this limit for its own Ads API, and n8n users run into the identical wall on Gmail, Sheets, Drive, and Calendar credentials, because n8n's Google nodes authenticate through the same OAuth 2.0 flow. Move the app's publishing status to "In production," and the seven-day clock goes away. From there, a refresh token stays good until you revoke it or leave it unused for a long stretch.
It isn't broken. It's built that way on purpose
The seven-day limit is Google's guardrail for apps nobody has reviewed. Anyone can spin up a Cloud project, paste a Client ID into n8n, and start pulling their own Gmail data in under ten minutes. If that convenience lasted forever with zero oversight, it would double as a decent phishing setup. Google's trade is this: an unreviewed app works fine, but only for a short list of test users you add by hand, and only for a week at a stretch before it has to prove someone's still paying attention.
How to check if this is actually your problem
Open Google Cloud Console, go to APIs & Services, then OAuth consent screen, and look at the Publishing status field. If it reads Testing, you've found it. Worth confirming before you chase anything else. n8n throws the same invalid_grant error whether a token expired from the seven-day limit or got revoked for some unrelated reason, and the two look identical in the n8n logs.
Does clicking "publish app" actually fix it?
Usually, but not always instantly, and not for every scope. Google sorts OAuth scopes into non-sensitive, sensitive, and restricted. Non-sensitive ones publish with no review at all. Gmail, Sheets, Drive, and Calendar mostly fall into sensitive or restricted territory, so publishing triggers Google's verification process: you write a justification for each scope, record a demo video of the consent flow, and restricted scopes add a security assessment on top. Google doesn't publish a fixed turnaround for any of this, and the clock pauses every time a reviewer asks a follow-up question. If your n8n setup touches a restricted scope like full Gmail access, plan on more than a week before it clears.
A "Testing" OAuth app isn't a smaller version of a finished one. It's a different product with a built-in expiration date.
If publishing doesn't fix it, rebuild the credential
One wrinkle comes up often enough to flag. Publishing to production sometimes doesn't help if n8n is still holding a token issued under the old Testing-era client. Builders in the n8n community have reported the fix only sticking after generating a brand-new OAuth Client ID and Secret in Google Cloud, dropping those into a fresh n8n credential, and running through the consent screen again from scratch. Reusing the existing credential and just re-authorizing isn't always enough — some of this ties to the client itself, not only to the project's publishing status.
What a silent failure like this actually costs
A seven-day death is worse than a loud crash, because nothing announces it. The trigger stops polling, so there's no failed execution sitting in the n8n log to notice — just an absence where new leads used to show up. We've written before about catching the n8n run that never happened, and the same heartbeat pattern applies here: a second scheduled workflow that checks whether the first one actually fired in the last 24 hours, and pages someone the moment it hasn't.
If you're running n8n credentials against Google services for anything client-facing (lead capture, invoice reminders, calendar sync), check the publishing status today. It's a two-minute look and it's usually the entire fix. If your scopes are sensitive enough to need Google's review, start that process now instead of after the third outage. We help businesses audit exactly this kind of thing as part of our AI and automation work, and a quick conversation is free either way.
— Cole
Sources
- Google Ads API — Common errors (documents the 7-day refresh token limit for apps in Testing status)
- Google Cloud — Submitting your app for verification
- n8n Community — "Anyone facing Google access blocked error?"
- LinkedIn post that prompted this piece — an automation builder's n8n auto-responder going dark for 7 days with zero executions, traced to the same Google OAuth Testing-mode limit