This is not exactly a new phenomenon. In the '80s - '90s, back in the days when you could call the duty forecaster and talk to him (foc) TAFs were regularly and obviously amended, sometines on an hourly basis, as a result of observations that differed from the forecast. In other words, TAFs were being re-written en masse to reflect the METARs being recieved.
With the vast - exponential - improvement in computer power and programming it sems hardly credible that forecasting has got worse since then. In fact I'd say that is an impossible accusation to justify.
We all (should) know that the tiniest variation in temp or pressure can have a dramatic effect on the weather, particularly the formation of fog which is probably the hardest thing to predict accurately.
I always suggest in a briefing that if a TAF is showing reduced vis and low cloud - eg PROB 30 TEMPO 4000 SCT006 - especially at night or towards dawn then to anticipate RVR0300 /// as a real possibility. This is often proved correct.
Maybe it is our expectations, as well as our incresed relianace on Met services - rather than our own Met knowledge and skills and common sense that are to blame?