Recently we implemented jumbo frames in our RAC environment, I have read a lot about jumbo frames and how they can be helpful in a cluster based environment but never actually experienced in a real time setup.
In one of the setups recently after migration , it was basically a hardware based migration, an upgrade to more advanced servers ,post migration database was slow and there were many Cluster related events. One of the main event being "gc cr request" and "gc cr congested" , these 2 events were in large count if we query gv$session_wait view.
There was no quick solution to this event and we tried to tune a few problematic sqls, which helped in reducing this event to some extent but most of the times the average CR block Receive Time was unusual then normal and so was average current block receive time.
Finally after a few days we had implemented JUMBO Frames in our rac environment, And just after implementing these, the average CR block Receive Time came down tremendously and also gc cr request and gc cr congested event were very very less.
So Jumbo frames are very much useful as i would say, although it is not recommended by support people.