I think this is definitely possible in TD, however how many lines you want and performance you expect or need may determine if you go about this via line sops, chops, and eventual conversion back to line sops, OR if you do a gpu/glsl approach.
Regardless of which way you go, I’d say the main stages here would be:
Take your depth pass, and do things like scale res down to something reasonable - IE row of pixels for each “scanline”, and set the number of rows to a number that seems aesthetically reasonable.
Add some form of turbulence to each row of pixels, you can do this in chop space for an easier workflow and r&d, but best to eventually port that back to top space if you can for performance reasons. It wouldn’t be a straight noise top though because in the video there’s clearly another mapping methodology at work, more 1d or row based than 2d projection based, if that makes sense.
Find a way to procedurally hide segments of each horizontal line that fit certain criteria, such as normal angle, or wrapping from end of last, to first of next, etc. the video shows a really nicely done “cutout” effect where things at different depths are separated and not connected. There are lots of ways to do this, both in glsl and nodes, but this will likely be the more complicated bit to get right.
Anyways, just some thoughts!