Oh, this looks cool 😎

Adaptive sampling is great, I meant something more deterministic. Like the rules we can define with the OTel tailsampling processor 🙏

I've seen engineers give up on tracing too soon after not finding the traces they are looking for. Deterministic sampling rules help a lot.

Now all I want is AWS managed and configurable tailsampling so that I can do always_on sampling on everything and send the traces to the OTLP endpoint from all workloads 🤓

Looks like the version attribute is still not on the plate (for standardization)? It helps in deployment tracking, identifying faulty deployment during an incident from the telemetry data etc.