We check the time spent to run benchmarks doesn't increase with the
number of iterations.