Not enough rep to comment, but worked on a similar setup recently. What do you mean by consumer “stops”. On rabbitmq server side, does the consumer still exist against the queue? You can check this on the dashboard.

This will tell you if your consumers really were stopped or if they’re up but not processing messages in your application.

If its the former, next clue would be checking for implicit channel closures (which may be autorecovered so you miss it happening?). Maybe set up a client/channel closure listener?

If its the latter, is it a threading issue? Are consumers processed on separate threads, and are they blocking their thread.

Also, do the logs on rabbitmq node show anything? Use named connections to pin it down. Let me know how you get on.