1)I would prefer to log to an Application Insights resource on Azure, but is this recommended?
Ans : You can definitely log to Application Insights. However, if you prefer not to use Azure resources, you can also log the data to a table in Dataverse. You could set up custom usage tables in Dataverse for this purpose.
The Dataverse Table is only recommended when the log volumes are very less - the volume is going to be high it is better/recommended to use App insights. However, Dataverse is cost efficient - because you are not using additional azure resource.
2) How can I manage the alerts that are fired based on the flow performance? For example, when a flow has failed several times.
Ans: Are you referring to Azure Application Insights or something more general?
If you're using Application Insights, you can set up an alert to check the flow failure count within the last hour. For example, if there are more than 15 failures (assuming that’s your threshold), you can trigger an alert.
Write a query some thing like below
PowerAutomateFlowEvents
| where Eventname == "FlowFailure"
| where timestamp > ago(1h)
| summarize failureCount = count() by bin(timestamp, 1h)
| where failureCount >= 15
If you're logging data in Dataverse, you can create a monitoring flow that checks the log table every hour, counts the number of failed rows, and sends an email alert if the failure count exceeds your threshold (like 15).
3) When should the terminate action be used? For example, if I have a parallel branch that runs when the previous action fails, do I need a terminate action, or should I let the flow end normally?
Ans : You can let the flow end normally in most cases. I typically use the "Terminate" action to stop the flow early without running any further steps. If one action fails, and you still want to continue with other actions because they’re not dependent on the failed one, you don’t need the "Terminate" action. The "Terminate" action should be used if you want to prevent unnecessary actions from running after a failure or produce incorrect outcomes after a failure
- Sathya Vijay