Tesla doesn't want to give all their secrets away too soon and it's a good bet it's not a friction clutch, but a positively locking one. One method that occurred to me was to use data from rotational sensors to perfectly synchronize the speed of the motor with that of the drive unit at the moment of engagement and only apply more than minimum torque once there was confirmation of engagement. A proper implementation could result in essentially zero wear, even on clutch capable of handling very high torques.
Disconnecting the motors removes the bearing friction of the motors from the equation (as well as any residual magnetic drag) while simultaneously allowing the most efficient motors to be used. My guess is one compelling factor was being able to maximize regenerative braking efficiency. Loaded trucks rely on their brakes much more often than passenger cars so optimizing regen efficiency would be key for driving in traffic and maximizing regen potential may have been necessary for descending mountain grades while fully loaded without use of friction brakes. The shape and timing of the magnetic fields of the motors designed to maximize regen efficiency will vary a bit from that of a motor designed to maximize efficiency of power delivery.
It makes good sense that tailoring different motors to excel at different tasks, and being able to uncouple them at will, is the kind of thinking that turns the impossible and impractical into tomorrow's trucking norm. The short-term thinking instigated by the corporate cost-cutting of the 1980's is what made this kind of innovation go away. We should all be thankful that the world has Elon and his teams who love to innovate better ways to do things. This is why I invest in TSLA. It's not our fault that the rest of the investment community doesn't understand what this means for future demand and profits.
Also hints at what's in store for Cybertruck! If they utilize the Semi's system for it (or something along these lines), then 500 miles definitely seems possible!