What exactly do ccd with kinematic objects mean?

Bowsing through release notes I found the following:

This version of PhysX allows for Continuous Collision Detection (CCD) on kinematic objects, which allows for accurate collisions between very fast moving rigid bodies

What does it mean in practical terms?

Does it mean I no longer need to trace for missing overlaps between position of objects between two frames if it’s choppy frame rate ?

Like in Robo recall gif what’s different that’d not work in earlier versions ?
Otherwise enabling ccd on the simulated body allowed it to collide without tunneling anyway in the earlier versions too. So what’s new?