~~~~~~~~~
Sorry, but, based on knowledge of the post-processing that is applied to Google Earth data, I am very confident that those are minor differences (faint lines resulting from combining vessel scans with other data) that were enhanced by the "shadowing" filter used to give the image a 3-D appearance.
First of all, side-scan sonar looks out to the sides - always leaving a "dead zone" line directly under the track where the instrument was towed. This website explains and illustrates the phenomenon -- and explains how closely-spaced scan tracks can be used to fill in the data gap.
If the "gap-filling" (by scan overlap or digital interpolation) is imperfect, faint lines remain in the resulting digital image.
Rather than embark on a (even more) pedantic discussion of gradient convolution matrix math, I did a little demo:
I created two identical pairs of orthogonal lines which were only faintly different in brightness from the background. Then I applied a shadowing algorithm (like used by Google Earth) to the pair at lower right. Voila!
~~~~~~~~~
FWIW, such algorithms are quite useful for analyzing overhead imagery. My son bought a big stack of WWII (grayscale) BDA (Bomb Damage Assessment) photos of German targets at a flea market. On one photo of a bombed-out German airfield, I noticed faint lines in some adjacent grassy fields.
Application of a similar "shadowing convolution" algorithm turned those faint lines into clearly-visible pairs of "ruts" leading to seventeen (clearly visible due to their "shadows") aircraft that had taxiied out into the woods -- and escaped the bombing of the airfield...
As to modern applications for such algorithms, "deponent further sayeth naught"... '-)
~~~~~~~~~
But you were correct re the Google Earth pattern:"it’s certainly not a natural structure."...