People have shown in the past (and M. Böck & I have really driven home for Cyg X-1 in 2011 & 2014) that power spectra (ways to measure contributions to overall variability at different timescales) are highly energy dependent. We confirm that this trend is crucial at low energies.
This means also that when calculating coherence and lags, we cannot work with a broad reference band (lots of literature does!). So this is going to be fun - people will need to change their methods.