Correct behavior is Each request is routed to ONE ARBITRARY CloudHub worker out of ALL three CloudHub workers

* Mule Domain Project is ONLY available for customer-hosted Mule runtimes, but not for Anypoint Runtime Fabric
* Mule domain project is available for Hybrid and Private Cloud (PCE). Rest all provide application isolation and can't support domain project.
What is Mule Domain Project?
* A Mule Domain Project is implemented to configure the resources that are shared among different projects. These resources can be used by all the projects associated with this domain. Mule applications can be associated with only one domain, but a domain can be associated with multiple projects. Shared resources allow multiple development teams to work in parallel using the same set of reusable connectors. Defining these connectors as shared resources at the domain level allows the team to: - Expose multiple services within the domain through the same port. - Share the connection to persistent storage. - Share services between apps through a well-defined interface. - Ensure consistency between apps upon any changes because the configuration is only set in one place.
* Use domains Project to share the same host and port among multiple projects. You can declare the http connector within a domain project and associate the domain project with other projects. Doing this also allows to control thread settings, keystore configurations, time outs for all the requests made within multiple applications. You may think that one can also achieve this by duplicating the http connector configuration across all the applications. But, doing this may pose a nightmare if you have to make a change and redeploy all the applications.
* If you use connector configuration in the domain and let all the applications use the new domain instead of a default domain, you will maintain only one copy of the http connector configuration. Any changes will require only the domain to the redeployed instead of all the applications.
You can start using domains in only three steps:
1) Create a Mule Domain project
2) Create the global connector configurations which needs to be shared across the applications inside the Mule Domain project
3) Modify the value of domain in mule-deploy.properties file of the applications

Correct answer is Use a Batch job scope to bulk insert records into the database
* Batch Job is most efficient way to manage millions of records.
A few points to note here are as follows :
Reliability: If you want reliabilty while processing the records, i.e should the processing survive a runtime crash or other unhappy scenarios, and when restarted process all the remaining records, if yes then go for batch as it uses persistent queues.
Error Handling: In Parallel for each an error in a particular route will stop processing the remaining records in that route and in such case you'd need to handle it using on error continue, batch process does not stop during such error instead you can have a step for failures and have a dedicated handling in it.
Memory footprint: Since question said that there are millions of records to process, parallel for each will aggregate all the processed records at the end and can possibly cause Out Of Memory.
Batch job instead provides a BatchResult in the on complete phase where you can get the count of failures and success. For huge file processing if order is not a concern definitely go ahead with Batch Job

An SLA for the upstream API CANNOT be provided.