The problem with complex, interconnect models like these are that you can’t just simply “fix” one thing and make it better. Let’s say you increase the resolution. While there are huge benefits to that, there are also potential downsides because every weather model has estimates built in and those may not be valid anymore at a different resolution.
It also depends on what is being verified. The modeling centers run and assess for quite a while before release. But if they are focused on, say, heights and temperature, they may not care as much if east coast cyclones are now not handled quite as well.