Hi Fabian,
I started modelling with Blender last year, amazing piece of software, keep it up! Not a Pro modeller yet but I think I can give you some insides based on my experience.
First question: I don't know how MSFS is managing the 3d models but I guess they work similar as in FSX (it should be even better handling more polygons and other matters for sure). Basically, it depends on you and how you manage yourself with your project, however there are some advantages/disadvantages. For example, let's say that you have a frame with a window in it, if the model is really complex and you need to have flexibility it is better to separate the objects so you can position them freely without even touching the edit mode in Blender, otherwise, if you have for example a screw that you need to move somewhere else in your frame then you need to enter the edit mode, select all the faces related to your screw and proceed with the relocation of it. That's not really efficient in terms of time plus you lose some abilities for positioning your model. However, if your model is low poly it isn't worth it because in this case separating the object will only increase the number of vertices in your model (consider the frame window example, 4 vertices for the window polygon and 4 extra vertices in the frame where the window is supposed to be attached), in this case it is better to assign different materials for your frame and the window and keep it as a unity.
There is one little drawback in Blender when separating objects and that is when baking AO or Normal maps. In this case, mixing all objects in one might help increase baking time (did this as an experiment), however, if you have a cage for each component it's better to go one by one in case one of them fails to achieve the desired results i.e. the normal maps. It will take more time but you have more control over the baking process.
Second question: I think there is a huge amount of information going around related to the amount of vertices but that depends on the end purpose of your model. 4 Vertices polygons are the standard everywhere as they keep it easier when working with edge loops, subdivision (turbo smooth), UV maps, etc. however that doesn't mean that you are not allowed to use triads from time to time or even N-Gons (polygons with more than 4 vertices) nevertheless I will avoid the last ones since you don't have a good control of the surface and they might produce unexpected distortions in your textures and can give you headaches for hours if you are not aware of it. So in summary, 4 Vertices polygons all the way, 3 Vertices polygons only when required, and N-Gons avoid them at all cost. You need to consider that at the end of the road, everything will be converted to triads.
Hope you find this useful and feel free to aks if something is not clear enough.
Regards,
Luis