Quicklooks are generated by power detecting, averaging and decimating in both azimuth and range directions by a configurable amount. IW and EW SLC product quicklooks are first de-burst and merged. the blue channel (B) representing an average of the absolute values of the two polarisations.
Better that they provide a real explanation to real users -- someone who knows what all this insider jargon means would not be reading the help page in the first place: "power detecting, averaging and decimating in both azimuth and range directions by a configurable amount. de-burst and merged"
Blue is either miscalculated or intentionally turned down.
I don't see how adding two reasonably ranging positive numbers from R and G could produce 99.6% of the pixels having a grayscale less than 15. They must have crammed values down to the low end with an extreme use of the 'curves' tool in gimp or counterpart. The histogram below is for S1A_EW_GRDM_1SDH_20150103T060735_20150103T060809_004003_004D21_249B.quicklook.png
But if this was just for the pleasing red and yellow optics, they could simply have taken blue identically equal to 1 on the grayscale and gotten something visually indistinguishable, at least on this image. Been there, done that.
But the 'average of the absolute values of the two polarisations' could be a valid re-normalization effort, the purpose being to tone down excesses in the red-green channels. Maybe in some images this takes out gratuitous non-uniformity and mitigates noisy pixels.
I had some doubts that the documented algorithm was used [for blue channel]
Maybe they will give you a free ride on the space shuttle if you alert them to this problem?
Surely they intended square root of sum of the squares rather than sum of absolute values. Then to calculate the overall mean of the whole image and its std deviation, to divide for the multiples of it. These would reside in some small range mostly at the low end, which could be stretched to [0,255] to assign to each pixel in the blue channel.
Finalize the file system and downloads by Q2 2015?!? Holy toledo, that should have been finalized -- using synthetic data -- on the day of the launch. Parallel processing, not serial. They knew exactly what the data would look like the day the design was finalized.