How do Select, SelectMany, and GroupBy differ, and how do you choose between them?

In production I account for cardinality and execution location. Flattening can multiply rows dramatically, and grouping in memory retains state proportional to the distinct keys and elements. For IQueryable, I verify translation and generated SQL because complex projections or groupings may execute differently across providers. I use stable keys rather than mutable display names and define how null or missing keys behave. If only aggregate values are needed, I calculate them without retaining complete groups where possible. Clear intermediate names or an explicit loop can be better than a deeply nested expression when the transformation includes validation, error handling, or several business rules.