In my experience this is down to four things:
1. Timeout values
2. Storage Unit Fragment Size
3. GRT Proxy
4. Version of NetBackup
For the timeouts it helps to set the client read timeouts on the Media Server and client (each DAG member) to be high - 1800 or 3600
The Storage Unit fragment size doesn't help for backups you have already done but i always use a Storage Unit Size of 5000MB on any disk or de-dupe storage unit that will be doing GRT backups (in fact i tend to use this size for any disk anyway!)
You can also try using the Media Server as the GRT proxy to see if that helps (client host properties - exchange section of all DAG members)
Finally always keep NBU up to date - there were hardcoded bpcd and bpdbm timeout changes in later versions of NBU to overcome just this issue - 7.1.0.3 had it first addressed and you say you are on 7.5 but worth making sure you are up to date and always keep application clients on the same level as the Media Servers
Hope this helps
