The 90d AF problem is real.
I've just spent about a week using a rental 90d body side by side with a 5dsR and it performed exactly the same as the one I own that is now at Canon repair. We'll see what that one looks like when it comes back.
As I've noted previously, I'm using both cameras with an EF600/4LIS II and either 1.4xii or 2xiii teleconverter. Maximum aperture with those TCs is f/5.6 and f/8, respectively, and that's at, or right next to, the maximum aperture where the AF system is rated to function. YMMV, depending on focal length and aperture of your lens.
My conclusion is based on comparison to the AF system in the 5dsR which I've been using for the past 5 years or so, making some hundreds of thousands of images and doing many microfocus adjustments (MFAs).
Compared to the 5dsR AF system, the 90d behaves oddly: when doing an MFA the same way I've done it for the 5dsR, a 5ds, 5dMarkII and a 7dII, the visible area of critical sharpness is very variable (the critically sharp plane can be easily seen to vary from front to back and everywhere in between on sets of 10 frames where I make slight manual changes to focus before hitting AF, in OneShot Mode and single point selected) and it doesn't respond to major changes in MFA setting (from +20 to -20 in 5 unit increments, the location of the critical sharpness plane remains in the same, variable, range). But, when used on live subjects, birds, it doesn't seem quite so bad (maybe that's just a consequence of it being pretty close at zero MFA and the depth of field covers it; if it was as far off as the 5dsR, which needs a 15 unit correction, it would be horrible.)
Where the 5dsR was hitting perfect focus time after time over 80 to 90% of 50 to 100 frames, the 90d might get 50% almost right. And that's with the benefit of a bit extra DOF from f/11 on most shots.
I can only conclude that the 90d has an inferior AF system and isn't up to handling my lenses the way I use them. I can also only hope that it is a software issue and a firmware update is in the works.