What is the practical difference between value types and reference types in C#, especially when assigning variables or passing arguments to methods?

I use this model to reason about ownership rather than relying on the simplified idea that classes are always mutable and structs are always cheap. Large structs can be expensive to copy, mutable structs can produce surprising copy-and-change behavior, and reference types can be designed as immutable values. in can avoid copying a large readonly struct, but the runtime may create defensive copies when members are not readonly. I would choose a type and parameter mode from the semantics the API needs, then verify performance before adding by-reference complexity. Across async or iterator boundaries, ref-like storage has additional lifetime restrictions because it cannot safely escape to the heap-backed state machine.