Portainer Swarm node labels and CPU limits for a low-traffic stack: which constraint actually controls placement?

I'm consolidating a low-traffic internal stack onto cheaper nodes in a Docker Swarm managed through Portainer. The plan is to label two small nodes (for example tier=low) and use placement constraints plus CPU/memory limits in the stack definition so these services never land on the larger nodes and never consume more than a fixed share of resources.

What I'm unsure about is how Portainer actually applies these settings when a stack is deployed through its UI or GitOps integration. Specifically, whether placement constraints defined in the compose file are preserved as-is, and whether the resource limits I set in Portainer's container/stack editor override or merge with what's in the compose file.

I'm also unclear on the failure behavior: if no node with the matching label has capacity, does the service stay pending, or does Swarm schedule it elsewhere?

My questions:

• When deploying a stack via Portainer, do compose-level deploy.placement.constraints and deploy.resources.limits take precedence over values set in the Portainer UI, or vice versa?
• Is there a supported way in Portainer to verify which labels a node actually carries before scheduling, beyond the Swarm node list?
• Are these behaviors consistent across recent Portainer CE/BE versions, or should I pin my testing to a specific release?