hi folks! i'd be interested to hear what people are doing to get a good understanding of how a network of processes/channels is performing

ie. where there is contention/bottlenecks etc

there is often talk of how channels being a reified thing makes them easy to inspect/monitor, but practice I find them quite opaque

have used buffer implementations that can report their current count, but that only gets you so far

am currently playing with a channel implementation that records put!/take! latency... but not convinced it's a good idea.

wip: https://github.com/ragnard/profiled-chan

usually end up adding specific monitoring in the processes instead but that's verbose and specific