Socket tcp problem

I have a raw socket application and there are two servers located on the back of a f5 virtual server.The vs is configured to have a default oneconnect profile and tcp-lan-optimized profile. It works perfectly up to the point where some clients which are using persistent socket connections are added to the system. I have the network capture , please check attahcments for text export or wireshark screenshot of the capture. 10.87.35.79 is f5 and 10.20.12.35 is the client.Client always makes the same small request and the response is always around 10KB. After a certain point,I suspect a delay on the network, the client begins to send duplicate ACK s and as suspected after third DUP ACK a fast retransmission occurs.So the loss recovery works perfectly.But after this flow , when the client makes the next request , the very first packet of the response stream has PSH flag set .This causes error on th e clientside. It seems like an algoirthm is on action and it turns on the loss recovery mechanism. It is almost impossible for me to find which configuration item on the tcp-lan-optimized profile. Anyone has any idea?