A few ideas I've had that might be nice.
a setting for DoF (in feet or meters), then the aperture varies to try and achieve this for different focus distances & zoom... ideal for avoiding overly thin DoF vs insufficient background seperation for decent portaits of moving subjects (i.e. kids at weddings)
variable global sensor reset release: allows either rows or columns in sensor to start imaging sequentially, effectively creating a built in graduated sensitivity either horizontally or vertically. could come in handy with the 11-24
on sensor pixel binning (problematic for bayer matrix), for super sensitivity at a lower resolution.
FLIR connected to EVF or even on the rear screen to help wildlife photogs spot well camoflaged targets.
dark frame removal by using a single long frame taken well before use, then a minimum entropy algorithum (or similar) used to scale the frame to the image frame for ultimate correction, rather than the very basic method used right now.. could afford to be left on all the time then, and adds minimal random noise to all but the longest images.
maybe a scintillator plug in and a new range of X-Ray lenses!!.. just becasue it would be cool!
a built in accelerometer linked to software for automatic video anti jitter in camera.
of course we all expect ETTR, focus peaking, trap focus and so on, after all ML's had them for some time now.