Two of the questions in the initial message i would like to answer.
How important is METAR parsing vs. offline weather modelling for you?
As i fly a jetliner, going a couple of hundreds of miles from one to the next airport, i always use live metar. See the clouds change below me. Be surprised by the deteriorating weather at the arrival airport during descent.
On the other hand there are people fling in a small area wanting to create an interesting scenario i.e. GA pilots or fighter pilots.
So i think both possibilities should exist. I'd rather have them very separated, not necessarily engine-wise but control-wise i mean. I myself need nothing but what the live metar gives me. But some people might find the real weather to difficult to fly, so i can imagine just one slider for them: realistic to easy. Easy would cut off the sharp edges of the real weather. There is no need to bother them by a daunting set of settings. Make that separate and as complex as needed for the ones who want to create their own scenario.
Are there weather effects which are currently missing but should be modelled better?
There are 3 things that would really enhance the feel of realism for me.
1) Get a rough ride when the live weather is rough.
Currently wind variation (at airports where it is in live metar) seem to be only in the layer up to about 200 ft. I understand that metar information is limited and cannot be applied realistic to the higher air layers. On the other hand getting rough weather only in the last seconds of landing is also unrealistic. So i think making some educated guesses about what might be aloft is better than nothing. I have seen clouds at fl 300. That must also be an educated guess?
Maybe something like this can be done:
From a single metar, let the wind variation work higher, maybe up to a 1000 ft or even better 2000 ft, so a full final can be flown in rough winds.
If taking surrounding metars into account, one could lift the level of wind variation even higher, depending on how many surrounding metars also have wind variation in them.
If the latter is the case and the engine decides to draw cloud columns all the way to fl 300, then it might be plausible that the variable winds are also taken up to that height.
2) High altitude visibility
The standard visibility aloft seems to be fixed set on something like 25 mi. This makes you always fly in fog at fl 300. My experience in real flight is that often one can see much further at that altitude. Maybe here also an educated guess about aloft visibility can be made. Possibly based on pressure and humidity in the metar below.
For now i just raised the fixed visibility aloft a bit, so i do not fly in fog all the time. But that is also not realistic, to always have good visibility aloft.
3) Bigger 3D cloud field
With 3D cloud range at max (45 mi) and i fly at fl 300 and look down, i see a rather small football field of clouds below me at maybe fl 50. Not only does this look strange, but it also limits my options for aloft visibility settings. If i set visibility further than 45 mi i can look beyond the cloud field to ground which i should not see. So i limited the visibility aloft to 45 mi, which is still a quite short distance.
I am aware that 3D clouds eat gpu. So i was thinking, how to get more clouds with less gpu. Maybe there is a possibility to give them LOD ranges (or just couple them to current LOD). In that case you get nice clouds when flying in them. And coverage far away when flying high.
I hope some fg weather experts can use these thoughts (or have better ideas of their own) to forge a better live metar sky.
Kind regards, Vincent