Why does Autopilot require human supervision? Because it's not finished yet. Human drivers still need to correct the computer's mistakes, and few humans have the reaction time of a race car driver.
I don't know exactly how fast computers will be able to drive eventually. I do know that optimized software running on silicon hardware is hella faster than human wetware. Probably the max speed of robocars will be limited by things like tire friction (to stay on the road) rather than robodriver reaction time.
Currently the limit is mainly the range of the sensors, not computer speed or if anything is "finished" yet.
If you can't see far enough ahead to stop or otherwise take action safely, even if your reaction time is
zero then you can't drive that speed.
Fast computers don't magically let the cameras see further.
The conti radar had a range of 160/525 meters/feet. The main forward camera (there's a reason they call it main) is only 150 meters (492 feet).
So a drop in max speed the system could safely operate at wasn't surprising.
Nor is the fact the "increase" rumored is only to 85 mph, not the 90 radar gave you. Math is:
At 90 mph the car travels 160 meters in ~4 (3.97678) seconds. At 80 mph (the vision max when released) a car travels 150 meters in...~4 seconds... (4.19426 specifically). At 85 mph a car travels 150 meters in ~4 seconds (3.94753).
Meaning 80 was still a bit conservative with vision compared to radar, giving you over 2 10ths more reaction time.
But at 85 you're basically "on par" in terms of how much time you have to reach the end of sensor range, or at least within a couple hundreds of a second difference.
Significantly north of 90, the car can't see far enough ahead to be able to stop in time- so no, the car can't drive at any speed it can keep on the road at- no matter how fast the computer gets- unless the range of the cameras is significantly improved.