[Still regarding comment #1 - I've also updated the description in the RFE]
Reply to comment #2:
The classification is the same, TCP, but there is no classifier involved. It is strictly forwarding based on a SFF, i.e. looking up the Service Path Header and deciding the next hop. Somewhere in port-chain #4, before reaching the TCP Optimizer function, a different classification criteria was matched in comparison to port-chain #1, which made traffic go into a different SFP, consequently (potentially) getting processed by different SFs, even if somewhere in the middle the same SF instance processes the same kind of traffic. Flows are encapsulated and isolated by different service path headers, so next hops are no longer only dependent on classification.
[Still regarding comment #1 - I've also updated the description in the RFE]
Reply to comment #2:
The classification is the same, TCP, but there is no classifier involved. It is strictly forwarding based on a SFF, i.e. looking up the Service Path Header and deciding the next hop. Somewhere in port-chain #4, before reaching the TCP Optimizer function, a different classification criteria was matched in comparison to port-chain #1, which made traffic go into a different SFP, consequently (potentially) getting processed by different SFs, even if somewhere in the middle the same SF instance processes the same kind of traffic. Flows are encapsulated and isolated by different service path headers, so next hops are no longer only dependent on classification.