There is a defect in NFA 9.3.6 which can cause problems where v5 or v9 Netflow can trigger NFA to sense bogus "reboots". This can cause SNMP profiles to not stay assigned to a device, spotty data, and/or no data.
In the harvester.routers table, you may see devices switch back and forth from PollState='Mapped' to PollState='RebootRefresh'.
A combination of issues in the code where v5 flows with multiple engine ID's were not being uniquely processed and v9 flows were using "FlowSequence" reboot detection versus "SystemUptime" reboot detection which is not efficient for some devices.
**This will need to be applied to each Harvester Server**