LINQ performance: when to avoid it_

For tight loops over contiguous data, Span<T> avoids heap allocations entirely. Combined with stackalloc for small buffers you can write high-performance code that is still readable and safe.

Counterpoint: unless you are writing game engines or trading systems, LINQ allocations are noise compared to network I/O, database calls, and JSON serialization. Optimize the actual bottleneck, not the pretty query syntax.

Every LINQ chain allocates enumerator objects on the heap. In a hot path called thousands of times per second this matters. Use BenchmarkDotNet to measure before assuming LINQ is the bottleneck, but know that for loops over arrays are often 5-10x faster.

The sneaky LINQ performance trap in C# is mixing EF Core queryables with in-memory LINQ. If you call .ToList() too early, you pull the entire table into memory and filter it in C#. Use EF Core's async queryable methods and only materialize at the last step.