How to prove safety of an Advanced Driver Assistance System (ADAS) or Automated Driving System (ADS)? Statistical validation in road traffic as used for simple ADAS quickly reaches the limits of feasibility for complex ADAS and ADS featuring SAE level 3 or higher. Scenario-based testing both in real prototype vehicles and in simulation offers an approach to tackle this challenge. In this process, the events in road traffic are described by logical scenarios based on parameters. This space defined by scenario parameters has to be explored in simulation, while the simulation itself is validated by comparison with real test drives. However, even in virtual driving tests that can be processed automatically the number of possible combinations of scenario parameters quickly gets too large when trying to test the entire parameter space. This paper presents a test case sampling method that iteratively explores the boundary between critical and non-critical test cases in the space of scenario parameters. New test cases are created near the suspected boundary, increasing the density of test cases in this region of the parameter space. This allows to characterize the parameter space with significantly fewer test cases to be simulated. The knowledge about critical limits of the system under test can be used for comparison of different software and hardware, continuous integration approaches as well as statements about the occurrence probability of critical behavior in road traffic. In this paper, the test case sampling method is explained and exemplarily evaluated for virtual validation of an Adaptive Cruise Control model in a target-cut-in scenario.