I've debugged a similar problem once. For me, my memory issue was caused by hung threads which would continue to consume resources. You may want to log the start and end of your request handlers, and make sure that all of your requests are getting to the end (cloudwatch insights queries could help count this). You could also create a second sidecar that is not receiving any load, and check if the memory is still increasing (of course if so, then there's some independent issue -- like possibly not your python app).