I want to hear people's opinion on this one.
I'm making asset packs to sell on digital stores like, for example, fab (previously unreal marketplace).
And one thing I always struggle with it to decide how much stuff should I pack into atlases?
On one hand - atlases save on draw calls, which is, obviously, good for optimization. On the other hand - I always have this question in the back of my mind: "but what if whoever buys the asset pack just wants some things from it, not all, so the rest of the texture space on the atlases will be just wasted".
I know that unreal, for example, has tools to pack different texture sets into atlases automatically. So in theory, if I make my props with separate texture sets, users then could pick whatever they need and merge them into atlases by using those tools.
So, what do ya'll think? Should I just not bother with thinking what and how many props I should pack into atlases and what should have separate texture sets?
Edit: Just to clarify what I mean exactly, here's an example. In the scene I'm working on currently there's a side table that has a reed diffuser, a small bowl, a book and a magnifying glass on it. All those props are tiny, and have 256p textures with average texel density of 11.5. They also don't have a common theme between them to justify packing them into an atlas, like I did, for instance, with flower pots. Having so many low res texture sets, however, increases draw calls for the scene. This is my dilemma constantly

Replies
Edit: Found it. Not as expansive as I remembered, but maybe it will give you some additional points of view.
https://polycount.com/discussion/comment/2799361
unreal won't batch across different meshes so trying to save drawcalls like that is misguided (yes, i know there are caveats)