I hate bumping an old post, but I am seeing the EXACT same problem as this thread is talking about.
My Bahtinov mask is showing the exact same "bananas". I figured this was a problem with my mask that I 3D printed. It looked clean, but maybe it wasn't sort of thing. But no matter how much I zoomed in, looked at other stars in the field, and looked at varying brightness on different gains, and they all look the same, bananas.
Then, while researching, I ran across this post and other information. One poster above said that "maybe you blue information from clouds is causing a second spike on a different focal plane", and then I thought about how Double EDs worked. They combine Red and Blue into the same focal plane. The whole point of the mask is to diverge the lines in the interference pattern according to focal plane. If a Doublet moves Red and Blue to one plane, and keeps the Green in another, then you would see a "U" shape across the rainbow, no?
This is my question, when using a mid-ranged focal length (F5-F6) and a super small pixel size (2.9um or smaller), can you see the "rainbow" in the form of a banana on a a B/W sensor? Being B/W, you wouldn't see the colors, just a white swipe of banana.
I am having poor skies for the next week or so, I can't wait to figure this out, it is bugging me. My thoughts are I can test with a narrow band filter. If I can bring that to focus (much like the image above), would that prove my point. Any information or anyone who could test/corroborate my theory, would be appreciated.
If you are wanting SOME proof of my idea, go check out the Bahtinov page on Wikipedia, the example image, if you zoom in, you will see the rainbows near the center of the image on the spikes, just like my theory postulates.