Use Trigger node action in business-central after failed async node execution leads to incorrect process status

The trigger method creates a new NodeInstance for the triggered node, which is still active when the process reaches the end node.

Manually deleting this NodeInstance using the Admin REST API before triggering the node results in correct behavior at process end:

curl -X DELETE "http://localhost:8080/kie-server/services/rest/server/admin/containers/wih-project_1.2/processes/instances/4/nodeinstances/1" -H "accept: application/json"

Caveats: This requires knowledge of the internals of the process instance (the NodeInstanceId) and the operation cannot be triggered from within business-central.

Using the Admin REST API, the issue can be avoided by executing the re-trigger endpoint instead of the trigger endpoint (requires knowledge about NodeInstanceId).

Using the attached reproducer, this can be reproduced either in business-central, or just by calling the matching Admin REST API endpoints (see script inside the reproducer). Here are the steps for the business-central approach:

1. Unzip case03091622-reproducer.zip and import the bc/wih-project project
2. Build & Deploy
3. Start an instance of the flow process without defining a value for the flag variable
4. The process execution will result in an exception during the async WIH execution
5. Define a value for the flag variable
6. Navigate to the process diagram, select the CustomWorkItemHandler node
7. CLick Trigger in the Node Actions box