My question relates to OUTBOUND SMTP email

If I have a connection document from a mail server (different NNN) to the HUB cluster what is the behaviour of the connection?   So you're using a cluster in the destination server name of the connection doc.  This actually creates 2 connections in the internal routing tables.  

1. Does it round robin outbound smtp email between the 2 hub servers in the cluster - if it doesn't how is the decision made for which HUB server to send to?

NRPC transfer to that server will be to a primary hub.  Since you are using a cluster name in the destination server of the connection doc, that means the connections will be equal cost.  In that case the lowest alphabetical destination server name will be chosen as primary.  hub1 is lower than hub2.

2. Does it automatically detect failover - i.e. hub1 is down so I'll send to hub2  

Yes, will failover to hub2 if hub1 is down.

3. Is there any load balancing (for outbound smtp) based on the availability index of the cluster as you would have if connecting to a database.

No.

4. If after a failover dose the failed server automatically become available again for outbound smtp routing.  

Yes, once the dynamic cost increase for the failed server has been cleared.  By default this happens on a 60 minute interval.  So, in sometime less than 60 minutes it will check hub1 again and transfer there if available.

Note my connection document from the mail server would be point to the cluster name of the hub servers for the task 'Mail Routing' (not a connecting document for each HUB server)

Thanks

Donald