What you're seeing here is simple light dispersion, as seen in a normal rainbow. White light (containing all wavelengths) refracts off of a curved surface such as a glass prism or a raindrop in a cloud, and the different wavelengths(B, G, R, IR, etc.) are bent and become visible at what appears to be different positions.
Trails from planes, called "Aerodynamic Contrails", display the same phenomenon. The pressure of the air passing over the fuselage apparently causes the temperature to drop, and in such a small area surrounded by warmth this causes moisture to condense in these cloud trails. The light disersion is in essence the same, then, as a rainbow seen in a rain shower.
This is an amazing picture though, as capturing these even with normal visible light photography on aircraft seems to be quite rare!
Update: You pointed out that the trails are different (the actual plane appears in different locations in the RGB image, in line with the corresponding color/band presumably). I believe that the effect really is the contrails and light dispersion, but that the different band sensors on the Sentinel instrument are capturing from slightly different positions/times. A fast-moving object such as an aircraft will then be moving along its path on different bands, and the trails/aircrafts in the individual band images will be moving laterally in the inverse direction from the Sentinel's satellite motion.
From the Sentinel-2 MSI sensor technical documentation, sourced here:
"Two distinct arrays of 12 detectors mounted on each focal plane covering VNIR and SWIR channels respectively. The 12 detectors on each focal plane are in a staggered configuration to cover the entire field of view."
The follow-up diagrams show what I believe to be the case, that for normal Earth observation at non-elevated features there would be no effect caused by this separation, but a unique feature such as aircraft would show up in this way.
Here are sources on dispersion and photographing contrails: