Why Power Automate flows shut down after 14 days of sustained throttling, how to recognize it, and how to reactivate the flow cleanly.
When a flow in Power Automate repeatedly hits throttling limits over an extended period, Microsoft eventually steps in: the flow is turned off automatically. This is not a coincidence or a platform glitch, but a documented safeguard meant to prevent a single flow from permanently blocking capacity without anyone intervening. For flow creators, it still often feels like a shock moment, especially when the flow is business-critical and suddenly shows as "Off" without a clear error message in the inbox. As of: July 2026.
This FAQ explains exactly when the automatic shutdown after sustained throttling kicks in, how you can tell whether your flow is affected, and how to reactivate it cleanly afterward without the problem immediately repeating.
Power Automate sets throughput limits for every flow and every connector, that is, how many actions or requests may be executed within a given time window. If a flow exceeds one of these limits, its execution is not stopped immediately but slowed down. According to the official documentation Limits for automated, scheduled, and instant flows flow activity automatically resumes once usage in the sliding time window drops back below the limit. It only becomes a problem when that doesn't happen and the flow stays constantly over the limits for an extended period. That's exactly when the 14-day rule kicks in.
Typical triggers for sustained throttling are too many actions per run, nested "Apply to each" loops, a trigger that fires too often, or a connector that responds to error code 429 ("Too Many Requests") because the underlying service limit has been reached. Such errors usually first show up as a noticeable delay in the flow run, long before the shutdown itself becomes visible.
Microsoft describes the rule in Understand platform limits and avoid getting throttled like this: All flows that are continuously throttled for 14 days are turned off. These flows can be reactivated at any time, but if performance doesn't improve, the platform will turn them off again. So the shutdown isn't a one-time event but a recurring mechanism as long as the underlying cause remains.
Important for context: this 14-day throttling deadline is not the only retention limit in Power Automate. The limits table for duration and retention limits lists several similarly named but distinct rules side by side:
Because these three cases are easy to mix up in practice, it's worth checking the notification email from Power Automate: it usually states clearly whether the reason for the shutdown was errors, throttling, or inactivity.
The most obvious immediate step is to edit the flow and save it again. Since throughput limits, according to Microsoft, apply to a single version of the flow definition, making a change and saving it again resets the time window. That buys you time, but it doesn't fix the actual root cause.
For a lasting solution, the following steps are worth considering:
$filter or $top, and remove unnecessary intermediate steps.In the Power Automate error code reference you'll also find specific guidance on individual timeout and throttling errors, in case you want to pinpoint the exact error source in the flow run history.
After the shutdown, a flow can be switched back on manually at any time. According to Turn a flow on or off a click on "Turn on" on the detail page, or "Turn on" via the context menu in the flow overview, is enough. What happens to events that piled up afterward depends on the trigger type:
If you don't want this backlog, according to the documentation the only option is to duplicate the flow and delete the original, which cleanly re-registers the trigger. That way you retain control over which data actually gets processed, instead of being surprised by a flood of catch-up runs. Anyone who wants to run such automations reliably long-term shouldn't ignore throttling warnings in the first place, but factor them into capacity planning early, for example as part of ongoing automation projects.
No. A single instance of throttling only slows the flow down temporarily, and it continues running automatically once usage drops back below the limit. The shutdown only happens once the flow stays continuously over the limits for 14 days.
Power Automate notifies flow owners and co-owners by email as soon as a flow is disabled. You can also check the analytics report on the flow's detail page to see how often the flow produced error code 429 or delayed executions over the past few days.
Yes, with a caveat. Because throughput limits apply per version of the flow definition, the time window restarts after every save. That buys short-term relief, but it doesn't fix the actual cause of the throttling if the flow structurally triggers too many actions.
It depends on the trigger type. Polling triggers such as the Recurrence trigger catch up on accumulated events once turned back on, while webhook triggers only process new events from the moment of reactivation onward and skip older events.
A Power Automate Process license provides the flow with dedicated capacity and higher limits, significantly reducing the risk of sustained throttling. According to Microsoft, it's the recommended measure for flows that structurally run close to the limits, but it doesn't replace optimizing inefficiently built flows.
Founder of NordFlux. Spent four years automating processes at enterprise scale at Dräger, and now brings that depth to the mid-market — pragmatic and with full data sovereignty.
Certifications
Why Power Automate throws error 429 and how you can specifically fix throttling through concurrency control and batching.
Why Power Automate approvals time out after 30 days and how the relay pattern of two flows reliably handles long-running approvals.
If you do not know the 14-day rule for continuous throttling, you only notice the shutdown once a process suddenly stops. NordFlux actively monitors your critical flows, catches throttling early, and prevents Power Automate from disabling the flow in the first place.