When starting a load and performance test, the VSM responds properly and honors the delay specified in the Virtual Service Image.
After some time, while the test is still running, the VSM stops considering the think time and sends the responses as fast as it can.
Steps to replicate the issue:
1. Deploy and start 2 Virtual Services.
2. Run Load test on VS and execute a transaction in the second VS.
3. At this point, both would be honoring the think time.
4. Now stop the second VS.
5. You would observe the first VS on which load testing was running would stop honoring the think time.