I know your problem, and I don't think windows or any multi-monitor software has fixed the issue yet:
Your portrait screen is much higher dpi than your landscape screen, so the pyhsical representation of your screens (where pixels on different screens are of different sizes) does not match the virtual representation of your screens (where all pixels are the same size)
Your actual desktop (the things on your desk) is a wide rectangle, because the widescreen rectangle has the same height as your portrait screen. Your eye sees one rectangular shape (ignoring the bezel).
I feel that if you moved the mouse two inches right it should move the same distance across the screen whether that's on the low-dpi screen or the high-dpi screen.
Your virtual desktop (what windows thinks you have) is a rotated L-shape, because all pixels are equal - there is no concept of physical size and the portrait screen just has more vertical pixels than the landscape screen.
The mouse is constrained to this L-Shape because it has one coordinate scale, based on pixels rather than physical location. At the point where you cross to the other screen, the DPI change affects both the cursor location but also the sensitivity; Suddently two inches of mouse travel equates to six inches of cursor movement instead of ten.
Until Windows (or some 3rd-party software) recognises the physical scale (dpi) of each screen, no mouse driver will have a way to take advantage of this info and compensate accordingly.
Windows cannot become resolution-independent fast enough, in my opinion