Currently, spec performance tests run only on the default SocketStream and ByteBufferNIO implementations. However, as we introduce performance optimizations to the driver’s networking, decoding, and encoding layers, these changes may impact both the default implementation and the Netty provider. To ensure comprehensive performance validation, we need to extend the performance test suite to run on Netty.

Running Netty-based performance tests helps detect unintended regressions and provides a baseline time series for performance validation. 

Since the majority of performance-related changes are surfaced with SocketStream and ByteBufferNIO, which are already tested on every commit, running Netty performance tests on each commit is unnecessary. These implementations provide sufficient early detection of regressions for most optimizations. However, Netty should be tested occasionally to ensure that performance changes do not introduce regressions specific to that provider.