diff --git a/abra b/abra
index f918ba6..8852970 100755
--- a/abra
+++ b/abra
@@ -802,10 +802,16 @@ ensure_stack_deployed() {
 
   local -a UNHEALTHY # mapping
   local -a HEALTHY # mapping
+  local -a UNKNOWN # mapping
 
   services=$(docker stack services "${STACK_NAME}" --format "{{.ID}}" | tr '\n' ' ')
   debug "Detected the following service IDs: $(echo $services)"
 
+  without_health_checks = $(docker inspect --format)
+
+  # TODO: this tells you if shit is missing
+  docker inspect --format "{{ json .State }}" traefik_app.1.rnpar574ozsn6oi2ouo5s84jn | jq "try(.Health.Status // \"missing\")"
+
   while [ ! "${#HEALTHY[@]}" -eq "$(( ${#SERVICES[@]} + 1 ))" ]; do
     debug "Number of services expected to be healthy: $(( ${#SERVICES[@]} + 1 ))"
     debug "Number of services actually healthy: ${#HEALTHY[@]}"
diff --git a/tests/apps/works/compose.yml b/tests/apps/works/compose.yml
index af454eb..8a3e189 100644
--- a/tests/apps/works/compose.yml
+++ b/tests/apps/works/compose.yml
@@ -1,7 +1,13 @@
 ---
+# The goal of this compose file is to have a testing ground for understanding
+# what cases we need to handle to get stable deployments. For that, we need to
+# work with healthchecks and deploy configurations quite closely. If you run
+# the `make symlink` target then this will be loaded into a "fake" app on your
+# local machine which you can deploy with `abra`.
+
 version: "3.8"
 services:
-  r1:
+  r1_should_work:
     image: redis:alpine
     deploy:
       update_config:
@@ -17,7 +23,8 @@ services:
       retries: 3
       start_period: 1s
       timeout: 3s
-  r2:
+
+  r2_broken_health_check:
     image: redis:alpine
     deploy:
       update_config:
@@ -29,8 +36,31 @@ services:
         max_attempts: 3
     healthcheck:
       test: foobar
-      # test: redis-cli ping
       interval: 2s
       retries: 3
       start_period: 1s
       timeout: 3s
+
+  r3_no_health_check:
+    image: redis:alpine
+    deploy:
+      update_config:
+        failure_action: rollback
+        order: start-first
+      rollback_config:
+        order: start-first
+      restart_policy:
+        max_attempts: 3
+
+  r4_disabled_health_check:
+    image: redis:alpine
+    deploy:
+      update_config:
+        failure_action: rollback
+        order: start-first
+      rollback_config:
+        order: start-first
+      restart_policy:
+        max_attempts: 3
+    healthcheck:
+      disable: true