While tuning ASC injections, I've seen that the ASC error signals sometimes by a lot depending on the thermal state.
Gabriele has mused upon this same thing while he changed CHARD Y controllers (alog 67981).

The following are movies of the eight arm ASC dofs error signals, starting at the beginning of a lock at 60 W last night and continuing for the next five hours.
Each PSD is of ten minutes of data, with 0.01 binwidth and 0.5 overlap.
There were no changes to the controllers during this lock.

I was hoping to observe changes to the L3 suspensions peaks one might expect from arm power increase as we thermalize.
Elenna has some nice comparisons of ASC OLGs for different arm powers (see CHARD P or DHARD Y),
where we see the modeled stiffening of the L3 spring with higher power for the HARD mode.
The SOFT suspension modes will soften with higher power.

The CHARD Y movie is probably the most interesting.  
There appears to be a broadband increase in the noise as we thermalize, except between 2.4 and 3 Hz.
Not sure why the broadband noise increases for this dof but not others.
Additionally, it seems like we can see a resonance moving from 1.2 to 1.4 Hz.
However, I don't think that could be the L3 Yaw mode, which Elenna's model shows to be nearer to 2.5 to 3 Hz. 

Each ASC movie exhibits a rising peak around 3.4 Hz.  This appears to be a L2L mode on the upper stages of the quads.  This explains why none of the usual models show this mode.
There are similar L2L resonances around 0.45 Hz, 1.0 Hz, and 2.0 Hz.
Unclear why this mode is rising as we thermalize.

In the end, these movies are difficult to interpret but still useful for identifying shifting resonances, cross-couplings, and rising modes.
The code is from Dan Brown and Austin Jennings, I moved it to 

/ligo/gitcommon/psd_movie

so I could so some further customizations.