This approach is different from the 2 that I suggested in the ticket. Can you write up why you think this is the best approach?

I also think that the repair method you suggested is more thorough and better, and my method is relatively simple, in order to quickly repair production problems.

My idea for repair is thread always first get client._lock, then get coordinator._lock. When the heartbeat thread uses 2 keys to call client.poll to read data, the performance of the main consumer thread is less affected. I have tested the performance.

I hope you can use a better way to solve the design problem of multi-threaded shared client.poll

quick question @dpkp : do you plan anytime soon a 2.0.2 release that could include this PR ?

Is there any plan to cut a release soon with this patch included?