Grid Bank Full - Cisco NCS5500

Hi everyone,
I am coming across the following error at work. Cisco TAC has given me a solution, but they have been taking forever to answer some questions about their solution, and I thought maybe someone here would know.
We have two NCS5500 devices running 6.6.3 that are giving out this error:
ipv6_rib[1232]: %L2-GRID-3-BANK_FULL : GRID POOL 2 TYPE GLIF BANK 1 is FULL

CISCO TAC RESPONSE:
"These are messages from grid which stores FEC of non-ECMP next-hops. If remote PE uses per ce label allocation, it can exhaust this resource on the NCS and we would see HW programming failed for the routes as well.

If this keeps happening, it is recommended to use label allocation per VRF for better usage of FEC resources. You can check resource usage by below commands:

sh contr npu resources fec location 

sh grid pool 2 bank 1 location 

Please let me know if you can make label allocation change."

The questions that I am trying to answer:
If we upgrade to 7.02 will this fix the issue by being more efficient?
If we do switch to per VRF do we have to do it network wide, or can we do it on just the two devices whose tables are full?

ANY INPUT IS GREATLY APPRECIATED!!!!!!