These combs have been found by using the spectral persistency, which is defined as the fraction of FFTs in which a "peak" is found at a given frequency bin. The whole dataset is divided into N pieces, and an FFT is computed for each of them. For each FFT an autoregressive estimation of the average spectrum, S_AR, is computed over the same data chunk and an equalised spectrum is obtained as |FFT|^2/S_AR. A peak is defined as a local maximum, above a threshold of sqrt(2.5) on the square root of the equalised spectra. For each frequency bin the persistency is a number going from 0 (no peak has been found at that bin in any of the equalized spectra) to 1 (a peak has been found in every equalised spectrum). See e.g. "P. Astone et al., Classical and Quantum Gravity 22 (2005) S1197" for more details.  
The frequency resolution is frequency dependent and goes from 6.22E-5 Hz at 10 Hz to 6.29E-4 Hz at 2048 Hz. 

The combs described below have been found using a threshold 0.3 on the persistency. By lowering the threshold several new potential combs appear, but it becomes more and more difficult to discriminate between real combs and random lines mimicking a comb. This must be studied more.

Even for the currently identified combs sometimes it is not easy to disentangle combs potentially sharing some lines (e.g. a line at 140 Hz could be member of both the 10 Hz comb and the 1 Hz comb). Some hints come from the persistency values which tend to be similar for lines belonging to the same comb (but this is not an "iron rule").

In the following we list the combs, giving the frequency of each line and the corresponding persistency.

*************
0.5 Hz separation
 
62.300000 62.800000 63.300000 66.300000 165.300058 355.800116 356.299958 357.299905 357.800011 358.300116 
0.728477 0.728477 0.728477 0.728477 0.695329 0.710949 0.713139 0.710949 0.718248 0.656204 

Note that these lines are calibration lines
*********

1 Hz separation
 
198.000000 199.000000 295.999903 297.999952  396.999889 398.000111 401.999887  497.000062  799.000000 1111.000093
0.719882 0.720862 0.709832 0.677058  0.692521 0.464681 0.560192  0.512074 0.568771 0.713695

Note here that lines from 6th to 9th have a smaller persistency. It is not clear if they really belong to the same comb as the other lines or not.
The code identifies one more line at 305.000000, but with an even smaller persistency (0.330448). 

*************
10 Hz separation

140.000000 190.000000 240.000000 290.000000 340.000000 490.000000 540.000000 640.000000 690.000000 1140.000000
0.473982 0.318739 0.410072 0.352595 0.402665 0.382043 0.340614 0.309614 0.322868 0.417313

Note some of these lines are separated by 50 Hz two-by-two. Not clear if this means something.

********************

9.999403 Hz separation with a bias of 9.999403/2=4.9997 Hz: f=n*9.999403+4.9997 Hz, n=10,11,12,13.

104.993425 114.992828 124.992217 134.991594
0.422721 0.438685 0.321993 0.392272

******************

9.9999867 Hz separation

109.999854 159.999824 169.999818 209.999798 259.999775 319.999751 359.999736 419.999715 459.999701 509.999685 559.999670 609.999656 659.999642
 709.999629 809.999603 859.999591 1159.999526
0.434610 0.462212 0.315287 0.469914 0.434708 0.416409 0.472263 0.449324 0.506779 0.507664 0.516091 0.367152 0.503504 0.490125 0.365272 0.37299
3 0.571138

******************

55.553551 Hz separation

55.553541 111.107093 166.660673 222.214358 277.767832 
0.359571 0.597977 0.593418 0.547032 0.588576 

These lines come from the Suspended Detection Bench 1 (SDB1) picomotors

***************