I agree things would be initially difficult, but I don't see how we could make a reader for less known disks without replicating their original drive mechanics

I agree things would be initially difficult, but I don't see how we could make a reader for less known disks without replicating their original drive mechanics as well, which is probably the hardest part. To me it would be more "scalable" to sample the magnetic field relative to the head position on a plane (on which we fixed the disk), then building a structure where each 2D element is mapped onto its position on the disk and assigned the value of the magnetic field. The result wouldn't be that different from this project pictures, but we could swap a 5.25 disk with a 3.5 one without changing a single screw in the reader. A further development could be reading more disks at once, provided their surface can be reached by the head. It would surely be useless while we have working disk drives, probably not 20 years from now...