The noise of a free-running oscillator close to the carrier is almost pure phase noise (see section 3 of http://www.designers-guide.org/Analysis/PLLnoise+jitter.pdf). This means that for low spectral frequencies, the variation over time of the cyclostationary noise in V/sqrt(Hz) will be almost perfectly proportional to the absolute value of the slope of the oscillator signal. With this knowledge, you should be able to reconstruct the time-varying power spectral density of a free-running oscillator at low spectral frequencies from its average power spectral density.