i whould say do what is most practical in the situation. if you need to, you can still join or seperate parts later
Blender
Community to share art made in Blender. https://www.blender.org/
I guess you could? though you can get some truly fucked topology in my limited experience of trying to merge stuff into a single mesh
with ctrl + j it should just join and not change the topology
@merthyr1831 It really depends what the final mesh is going to be used for. If you're just rendering in Blender multiple meshes are fine because the renderer will take care of most things and it's modular. Game? That gets complicated fast. For a prototype it may be OK but for production, you want to control overdraw and polys and bake the multiple meshes down to a game ready monolithic model. A tank in a game is going to be multiple meshes to some degree. Modular supports variations.
Of course there will be a few meshes per model, but I've seen people modelling both in the example of "just make a new mesh and sort it later" and "vigorously control the topology at every step of the process" which is also painful lol.
When 3D modelling relies so much on quads, it sure is hard to keep said quads from turning into ngons! just need more practice I suppose
@merthyr1831 It comes down to preference, how efficient you want to be and what best supports the end goal of the result. There's really no general purpose answer.
You can even have n-gons in subdivision surface models if it's a flat area.