C# is commonly seen as a high-level language with hard performance ceilings. When systems start slowing down, the reflex is often to blame the runtime and suggest rewriting critical parts in C, C++, or Rust.

But how far can you actually push C# before you need to consider a more fundamental redesign or re-platforming?

This session is for experienced C# developers who have tried the standard performance advice and need to go further. We'll briefly revisit allocation reduction through pooling and reuse, along with direct memory access using Span and related low-level primitives. Then we'll dig into more advanced techniques used in libraries such as the Azure SDK, RabbitMQ, and NServiceBus.

We'll look at removing abstraction overhead, shifting work from runtime to compile time, and using runtime and language features that don't get much attention. Just as importantly, we'll ask when redesign delivers more than another round of micro-optimization.

The goal is simple: understand how deep the rabbit hole goes, so you can push C# far beyond its perceived performance limits.