They don't mean the absolute real distance between dongle and mouse.

They mean the mouse communicates an absolute position (relative to some arbitrary 0,0 the mouse decides upon) instead of a relative direction.

Dongle can then take latest coord packet and diff it against previous coord packet to get a relative coord to pass via HID to the system.

If the RF packets are lost, some latency occurs but the dongle still has the previous mouse coord and can make a fairly accurate correction once a packet gets thru (get's from A to D, but might skip points B+C).

No, mice sensors are far from being SoTA linear encoders. They return approximate instantaneous movements somehow when read request is received. Mouse controller chips(USB or RF or Bluetooth) pack and report up the movements, that's it.

(psa: none of Chinese ADNS-2610 clones have the raw pixel output debug command. Maybe security implications or maybe something else, either way, mouse-as-microscope hacks don't work on sensors extracted from e-wastes)