I am using opencensus-python library to trace requests to my Python Flask application.

I have noticed that when using a StackdriverExporter instance, the memory usage of my app grows steadily for each traced request. When observed using memory_profiler, the memory usage grows roughly 700KB to 1MB per request. I have also noticed similar usage growth on GKE metrics dashboard when running the app on GKE.

All other factors being the same, if I use a LoggingExporter instance (instead of StackdriverExporter), the memory usage stays constant (as expected).

I have put together a sample app along with commands to repro this issue: https://gist.github.com/Amey-D/ebbf7acb77446e81998cf325c7737e73

Any ideas?