When we analyzed the trace we found that whenever there is a packet loss the backend server X.X.X.X sends a SACK but due to some reasons the SACK is not honored and Netscaler keep re-transmitting the same packet instead of sending the correct packet.
Netscaler compacting packets while retransmitting which is causing this connection hang. 

Frame No: 508977 ACK with sack blocks, it does not have prior sacked bytes we treat this as duplicate ack.

Frame No: 509000 One more duplicate ack, since outstanding is less than 4*mss we add this in to Early retransmission list, after 10ms NetScaler will do retransmission.

Frame No: 509016 Third duplicate FR is done. Netscaler retransmit only one packet, 509017 is sent (i.e 13 bytes)

Ideally other end should have accepted the packet and chop the extra bytes and send DSACK. Other end is sending Duplicate acks for new data sent by NS and DSACK for other duplicate retransmissions. Finally connection closed by the peer. 
 

Need to identify the device in between(by taking simulations traces) which is dropping the compacted packets and make that device accept compacted packets as per RFC compliance.
From NetScaler Prospect It's an expected behavior.

---

Problem Cause

The issue occurred as one of the devices in the network is not accepting the TCP compaction, which is an RFC compliance. we need to check with customer(after Engineering confirmation) on a new test to reproduce it and collect traces on NS, Back end server and middle boxes (like switch etc) which will tell us the reason for packet drop at the end devices as in ideal scenario it should be accepted.

When customer upgrades the Netscaler from an older version of the NetScaler, ICA Connections not working after NetScaler upgrade. Application access issues occured after upgrade from NS11.1 Build 51.26 to NS12.0 Build 53.6, apps not starting(stuck at starting session stage) or freezing/disconnecting after few second.