Currently we don't have any ability to drill down to the step level. Also the load happens for all steps unconditionally. What's going on is that if you specify 100 virtual users, Test Studio will start 100 separate threads, each representing one user and let each one run through the user profile (i.e. the HTTP traffic for that user scenario) as quickly as possible with appropriate pauses for the think times included in the user profile.
We don't do anything like detect the load during step 6 of the test. Since all the virtual users are running asynchronously each user can be at any random step at any point in time, but you will always have 100 active virtual users trying to interact with your application under test... up until the time runs out for the test.
I hope that helps clear up how our load testing works.
the Telerik team