Thanks for sharing the trigger JSON — that helps. Here’s what your automation is doing and why you may be seeing multiple runs. What your trigger means (Prefect 3.x): - It’s a reactive EventTrigger that fires whenever an event of type

cdf.publish_audited_tables

is received - It requires the event to have a related resource with

prefect.resource.role = flow-run

(i.e., typically when you emit the event from inside a flow run, Prefect includes that flow run as a related resource) -

threshold: 1

and

within: 0

means “fire immediately when the event arrives” Why you might get 3 runs ~30 minutes apart even if “the event is only emitted once”: 1) Event loop (most common) - If the flow that your automation runs also emits the same event (

cdf.publish_audited_tables

) when it runs (e.g., at the end), that new event will match the same automation and trigger again. If that flow takes ~30 minutes, you’ll see a new run ~30 minutes later, and so on. - This can happen even if you only intentionally emit the event once upstream — the triggered flow emitting it again creates the loop. 2) Multiple related flow-run resources attached to the single event - If the single event includes more than one related

flow-run

(e.g., parent and child runs), the automation can evaluate separately and fire for each unless you partition with

for_each

. 3) Action retries or duplicate automations (less common) - If the action (e.g., RunDeployment) could not be recorded as completed due to a transient error, the automations service may try again later. - Or there may be multiple automations or multiple actions configured that look similar. How to confirm the root cause: - Check the Automation’s Activity in the UI. For each firing, look at: - triggering_event.type and triggering_event.id — are they different events? - If they are different events of the same type at ~30 min intervals, it’s likely a feedback loop from the triggered flow. - Verify whether the flow that is being triggered emits

cdf.publish_audited_tables

itself. - Ensure there aren’t duplicate automations or multiple actions on the same automation. - You can also watch events live:

prefect events stream --format text

and filter for

cdf.publish_audited_tables

to see exactly when and how often they appear. Ways to make it fire only once (and avoid loops): - Partition the trigger per originating flow run: - Add

for_each: ["prefect.resource.id"]

- This evaluates the trigger separately per related flow run and prevents multiple firings for the same run if multiple matching events occur. - Avoid matching events emitted by the triggered flow: - If your triggered flow also emits

cdf.publish_audited_tables

, either: - Change the event name the triggered flow emits (e.g., make a distinct name), or - Add a distinguishing label to the original event (e.g.,

cdf.origin=upstream

) and add that to

match

so only upstream-originated events trigger:

"match": { "cdf.origin": "upstream" }

- Then ensure the triggered flow does not include that label when emitting. - Narrow related resources: - If your custom event unintentionally includes multiple

flow-run

related resources, only attach the one you need, or switch to

match

if the flow run is the primary resource emitting the event. Helpful docs: - Event triggers (Automations) in Prefect 3 - Automation schema reference -

prefect events

CLI help (for streaming): run

prefect events --help
  prefect events stream --help

If you can share: - A screenshot/snippet from the Automation’s Activity tab for the three firings (showing triggering_event.id and timestamps) - Whether the triggered flow emits

cdf.publish_audited_tables

- The action configuration (e.g., is it a single RunDeployment action)

…I can give you a precise fix. My hunch: adding

for_each: ["prefect.resource.id"]

and ensuring the triggered flow does not emit the same event will resolve this.