Mike Gorman

Thanks, @owais . I was able to reproduce the issue with this stand-alone test: ```python import time from opentelemetry.exporter.otlp.proto.grpc.trace_exporter import ( OTLPSpanExporter, ) from opentelemetry.sdk.resources import DEPLOYMENT_ENVIRONMENT from opentelemetry.sdk.resources import...

FYI - when I replace `OTLPSpanExporter` with `ConsoleSpanExporter`, I am not able to reproduce the issue (after 70 tests).

Restoring the original `OTLPSpanExporter` from my test but disabling the sampler, the issue now happens on every execution. So, previously, the reason it took "10 or 20 times" was because,...

To help illuminate what's going on during `shutdown`, I "monkey patched" `BatchSpanProcessor.shutdown` method in order to print out a traceback of each thread, to see what each thread is doing,...

Appears the worker thread is hung here: 

Thanks, all. I would like to get clarification if you are saying there is nothing we can do to influence this "process shutdown" behavior. For example, is there a timeout...

Thanks. Will await your thoughtful decision. But, just to be clear, we are talking about a process exiting **naturally** - no human or Ctrl+C involved. I don't know if SIGKILL...

That, unfortunately, will not work because we have at least 100 different scripts so there's no single point of entry and, thus, no single point of exit where we could...

I tested this with an explicit call to `provider.force_flush()` right before I expect the process to exit and all that did was add a 30-second delay. I next attempted to...

Alright! Thanks for the thoughtful response. We'll put our heads together here and decide how we want to proceed (or if we need any help). Everyone's desire to do the...