Unable to detect rollbacks during "docker stack deploy" in CI/CD

Description

We're running a docker stack deploy ... in CI/CD to deploy apps. When a service isn't updated successfully and a rollback is triggered, the deploy command still exits with code 0 indicating everything is fine. I haven't found a way to let it exit with another code, so the CI/CD pipeline can react on the deployment error (the rollback).

$ docker stack deploy --detach=false -c docker-compose.yml test-rollback
Updating service test-rollback_web (id: ximv0s6r8fgd17mdh6d100yuk)
overall progress: rolling back update: 2 out of 2 tasks
1/2: running   [==================================================>]
2/2: running   [==================================================>]
rollback: update rolled back due to failure or early termination of task 8jej8t73el7cgty3t5s4l6gwp
verify: Service ximv0s6r8fgd17mdh6d100yuk converged
$ echo $?
0

Used docker-compose (first deploy it with exit 0 in the healthcheck, then edit to exit 1 to get a rollback on the next deployment):

$ cat docker-compose.yml
services:
  web:
    image: nginx:alpine
    deploy:
      replicas: 2
      update_config:
        order: start-first
        failure_action: rollback
        monitor: 5s
        parallelism: 1
      restart_policy:
        condition: any
    healthcheck:
      # guaranteed to fail
      test: ["CMD", "sh", "-c", "exit 1"]
      interval: 2s
      timeout: 1s
      retries: 2

Start by tracing docker stack deploy --detach=false through its service update and rollback handling, focusing on how the command determines its exit status. Reproduce the behavior with the supplied docker-compose.yml healthcheck failure and verify that a rollback is reported through a nonzero status suitable for CI/CD.