The 2x factor appears when the compared cameras have the same photosites, as I already said.
If you assume the exact same pixel (size and read noise), say that of the 7DII then at the lowest acceptable shadows by PDR's definition SNR is 4.4 and a 5DS would show 4.7, a ratio of 1.07. In the disaster scenario with the 7DII at SNR =1, the 5DS would be at 1.25.
The whole demo was aimed at making the "FF is equivalent to M4/3 with two extra stops of exposure" false, as I already said too. In general the smaller sensor (or crop from the big one) will beat the larger sensor in SNR, as my cameras test and your example show.
Should we try it with more recent cameras, say EM-1II and a9? The a9 is better, the ratio of the two SNRs of 0.96 at PDR and of 0.78 in the disaster scenario.
In my opinion it is misleading to say that a smaller format will automatically have higher (or 2x) SNR than a larger one. That has virtually nothing to do with format size and all to do with the read noise (and other non-idealities) of the sensors and lenses involved. I'll leave it at that.
Jack