Keep error listener attached to job stream

Author: jwulf

.circleci/config.yml

@@ -33,7 +33,7 @@ jobs:
   test:
     docker:
       - image: circleci/node:14.2.0
-      - image: camunda/zeebe:0.23.2
+      - image: camunda/zeebe:0.24.2
     working_directory: ~/zeebe-client-node-js
     steps:
       - checkout
@@ -66,7 +66,7 @@ jobs:
   # Publishes to npm
   publish:
     docker:
-      - image: circleci/node:11.10.1
+      - image: circleci/node:14.2.0
     steps:
       - checkout
       - restore_cache:

DEVELOP.md

@@ -11,3 +11,11 @@ To publish a new package, update the `package.json` package version and push a c
 There is a test that checks that the client can reconnect to a rescheduled broker, or a broker that starts after the client application.
 
 It can be run with `npm run test:disconnection` and it is run automatically when committing changes. It requires Docker to be running on your machine.
+
+## Debug logging
+
+Try this for the insane level of trace logging:
+
+```
+DEBUG=grpc,worker GRPC_TRACE=all GRPC_VERBOSITY=DEBUG jest Worker-LongPoll --detectOpenHandles
+```