>You can insert identical-other-than-case files with some kernel calls and break everything when you try to modify them later with the more common file interfaces.
that doesn't seem like a realistic scenario, nor is it comparable to explaining to granny that "Cookie recipe.txt" and "cookie recipe.txt" are two separate files.
You just need to modify a directory with two different programs using different windows files apis (and create case-insensitive duplicates while doing so). How common that is will depend on how often an end user accidentally tries to create duplicate files (if that's infrequent then a completely case-sensitive system would also only cause issues infrequently), and in the proportion of programs using the different windows file apis.
I think you're right that we would expect that to happen only rarely, but it seems implausible to expect that there's an entire windows file api with zero usage.
Side-note: I don't think explaining "Cookie recipe.txt" and "cookie recipe.txt" being different would be thaat hard -- if they look different then they are different. As long as granny doesn't have to worry about zero-width whitespace or other garbage in her file names then that's a good enough rule of thumb.
that doesn't seem like a realistic scenario, nor is it comparable to explaining to granny that "Cookie recipe.txt" and "cookie recipe.txt" are two separate files.