We’ve had some amazing swings in temp forecasts recently, with the boundary between warm and cold air bouncing around in their models (I assume).
The fundamental problem is in giving a point estimate to what really should be a probability distribution. A more honest way would be to continually provide probabilities for VFR, IFR, and low IFR, but I guess that would be a struggle for people to make decisions based on that.
It's not temperature swings, as this isn't a new problem. It's that the process is too iterative for certain airports. It tries to change it's mind too fast when it should just let it ride for longer.
The GFA product does something similar to what you're describing. This replaced the FA (Area Forecast) product. But it suffers from a similar problem, or a lot of times, won't match what the actual TAF says. For area planning, the GFA product can be useful, but when you're having to make legally based weather calls, it's not refined enough, which as you said, can be a struggle to be useful.