While the iterative nature of forecasting has made the system more dynamic and reactive (in a good way) to events as they unfold, the issue I continuously have is that it's too reactive.
Some TAF stations will get updates more often than others. There's various reasons as to why that is, but airline use at those stations is certainly one of them. Unfortunately what ends up happening is that the updates will swing back and forth between LIFR and VFR...and then back again... in a matter of 4 hours. This makes it really hard to a) launch on a flight legally and/or b) launch legally only to have your weather collapse right in front of you. And then when you check the TAF again, sure enough, it's predicted it was going to be terrible again even though it said it would be fine two hours earlier for the same time period.
Local knowledge of course helps, but sometimes when you get burned so many times you end up just assuming the worst when you could have gone. All because of iterative negative reinforcement.