The amazing part is that it is common to have resolution of 1/100 of a Hertz on RF signals. Correcting frequency to lock to WWV is trivial bit of math.
Long ago, I had a 68HC11 processor with 512 bytes of EEPROM to store my program into, part of it was locking on to the speed of a production line, with a pulse that happened about every 15 inches (one revolution of the drive wheel) and adjusting timing in a simulated phase locked loop to get 1/10" counts. I had an DDS in there along with all the other code and it worked great. That part of it was about 50 bytes, if I recall correctly.
It's less impressive if you simply see the result, e.g. in ray traced images but what is done mathematically, is evaluate an expression of infinitely-nested luminance integrals.