The correct answer is C . A common cause of poor performance in legacy VPN architectures is hairpinning traffic through a central data center before it can reach cloud or internet destinations. This creates unnecessary distance, added latency, and congestion because the user’s traffic does not take the most direct path to the application. Instead, it is first forced back into the enterprise network, often through a VPN concentrator and a stack of centralized security appliances.

This design made more sense when applications mostly lived in corporate data centers. But once applications moved to the cloud and users became more distributed, the same architecture began creating serious user-experience problems. Zero Trust addresses this by allowing access to be enforced closer to the user and closer to the destination, rather than depending on centralized backhaul.

The other options are weaker answers. Split tunneling introduces visibility and control concerns, but it is not the main performance problem being tested here. Vendor throttling and IPSec version mismatch are not the common architectural cause. Therefore, the best answer is hairpinning cloud application traffic through a data center bottleneck .