Events in Cluster Operation are not forwarded after a Bad Gateway error from a load balancer

Current Situation

When several JOC Cockpit instances are operating as a cluster together with a load balancer and the load balancer returns a 504 "Bad Gateway" error then Events will no longer be forwarded to the browser of a User logged in over the Load Balancer's URL.
This means that the User will no longer be informed of changes in the state of a task until a refresh is forced by, for example, changing a tab or view.
At the same time, a User directly logged in to the URL of the Cluster member executing the task will be able to continue to follow its progress.
The 504 "Bad Gateway" error occurs when the Load Balancer's Read Timeout is exceeded.
The browser of a User logged in over the Load Balancer's URL then stops sending Event requests.

Desired Behaviour

The JOC Cockpit GUI should continue to send Event requests even after 504 responses.
The behaviour of JOC Cockpit GUI as a result of other potential responses should be optimized with the goal of making operation more robust.

Workaround

The occurrence of Bad Gateway errors can be avoided by increasing the Load Balancer's Read Timeout setting.