A manufacturer reports slow file transfers and ERP sync over a high-latency MPLS and internet mix, with packet captures showing mostly long-lived TCP sessions and window constraints. Loss is low and UDP voice is fine, so the team wants App-QoE optimization without touching voice. Which statement correctly describes TCP optimization benefits and the traffic it applies to?

The correct statement fits because long-lived TCP file transfer and ERP sync over high-latency MPLS and internet suffer from bandwidth-delay product limits, slow window growth and RTT-sensitive recovery, so window scaling, selective acknowledgment tuning and pacing directly raise throughput without altering voice policy. The forward error correction description fails by concept because FEC inserts parity to avoid retransmissions on lossy paths, yet the scenario explicitly states loss is low and voice is clean, so there is little loss to correct and window exhaustion remains unaddressed. The handshake proxy description fails by concept because edge connection splitting mainly cuts setup time for many brief SaaS transactions, whereas captures show mostly sustained sessions where setup cost is negligible and steady-state window behavior dominates. The payload compression description fails by concept because deduplication saves capacity on expensive constrained circuits, yet the bottleneck here is latency-induced window constraint rather than proven congestion or bandwidth exhaustion. Exam caveat: distinguish throughput-from-latency optimization versus loss remediation versus capacity saving, and read stem facts about session duration and loss before choosing. Practical check: confirm large RTT, constrained advertised window and flat throughput with capture analysis, then pilot TCP optimization on a single site pair and compare goodput and ERP sync time while confirming voice quality remains unchanged.