how Traffic Manager would distribute load in the following scenario

I'm wondering how Traffic Manager would distribute load in the following scenario:

I setup the same hosted service in two different datacenters (USNC and EUN for example) and create a performance policy in Traffic Manager that includes both. I understand the dynamic here, but what if I add a second hosted service in one of the two datacenters, USNC for example. How would a performance policy handle there being two hosted services in one of the datacenters?

In thinking about this I see two obvious options. The first is that it just uses a round robin approach for requests that are bound for the datacenter with two hosted services. The second option is that it pushes all traffic at the first hosted service listed for that datacenter.

I am really hoping it is NOT the latter as that would eliminate the possibility of scaling within a datacenter while also supporting other datacenters.

Any know how Traffic Manager actually handles this?

It is actually the latter - if one or more hosts are in the same DC it will always return the first hosted service in a performance policy.

In order to address that, one of the possibilities is to expand Traffic Manager to allow for nested load balancing policies, which would be more flexible, but such a feature is still not planned. Another possibility is to actually change the behavior to something like what you suggest - round robin between hosts in the same datacenter - but would need to be further evaluated by the team.

Could you please post this feature to the My Great Windows Azure Idea web site? I'd appreciate that so we can keep track of this request and allow other users to vote on it as well.