This ticket follows from . It was discovered that a significant amount of time, during checkout  = 9sec, publishing of  ITEM_CHECKED_OUT takes 7.6sec:

Furthermore, while monitoring threads within the pubsub container, it was observed that with each checkout, the number of threads increases up to 20. However, these threads do not get released afterward. The attempt to adjust the thread pool size to 30 did not impact the number of active threads. Even when enforcing a maximum execution time of 2 minutes, the threads did not free up as expected:

Additionally, the closing of sharedProducer has been moved to the finally block of the handler to ensure the proper closure of the shared producer, especially in scenarios where an issue might occur during sharedProducer.write().
Metrics collected using Rancher before and after implementing the mentioned changes for mod-pubsub show not significant differences:

before:

after:

Attached are logs from mod-pubsub for further investigation: