They observed that browsers don’t hide high-precision timers well enough (coarse time and jitter preserve Gaussian distribution), and suggest an alternative way of adding random delays to ruin timing measurements (section 6.1 in the paper).