Introduce profiling to find bottlenecks on throughput #660
Description
We should have some sort of profiling in place to find potential bottlenecks particularly around message processing.
We could explore integrating flamegraph and/or pprof or samply which has better MacOS support