You don't mention what versions of the hub you are using - that would be helpful. As would the platform. Also would be good to know how many get queues are being run on the upstream hub.
The message passing functionality in the hub is pretty bulletproof. I have literally thousands of such scenarios and can say that I've not run into any cases where the queues hung up for no external reason.
As a starting point, I would make sure that you are running the latest hub code CA Unified Infrastructure Management Hotfix Index - CA Technologies .
And look at it from a network standpoint - these hub to hub connections are persistent. Some firewalls will drop old connections regardless of whether they are moving data or not - Juniper I think does that by default. Or even the possibility of a duplicate IP causing packets to get lost.
-Garin