- Yes. Every subnet has limited resources and using resources puts that subnet closer to capacity. A lot of discussion on this topic is ongoing here
- As you can read in the thread I linked above, it is possible to avoid some of them, e.g. with a compute allocation.
- Depends on the scale and your requirements. You many even have to scale across multiple subnets if you reach a certain size. A subnet is inherently bound to what a single machine could accomplish on its own