Pulumi has a limitation when it comes to its AWSX Fragate service component as per these two issues:
If you have multiple ECS services, they will update one-by-one instead of in-parallel when continueBeforeSteadyState
flag is set to false
The flag is essential if you want your deployment build to actually fail when services aren't deployed successfully due to whatever reason
Disabling the flag will make everything update in-parallel but if one of the services is not healthy or rolled back then your CI will still not know
unless you build custom integrations with AWS EventBridge events
The following is a pulumi component resource that does what Pulumi should do by default and it behaves exactly as you would expect: