To Reproduce Steps to reproduce the behavior:
This problem was originally reported as a problem with inventory generation. During investigation it looked like broken HostnameMap is a cause:

2025-05-27 12:14:04.417233 | 566f0855-0029-11e0-4bd0-000000000009 |      FATAL | Generate ansible inventory | localhost | error={"changed": false, "error": "No IPs found for Rhel9Compute role on ctlplane network", "msg": "Error generating inventory for overcloud: No IPs found for Rhel9Compute role on ctlplane network", "success": false}

After applying workaround from https://access.redhat.com/solutions/7120161 (adding "provisioned: false" to source role and aligning counts) during FFU, generated baremetal-deployment.yaml has invalid HostnameMap: a mapping for destination role's hostname is not there. Instead a record is added for original role's instance.

This looks similar to the problem described and solved in tripleo-ansible patch attached to https://issues.redhat.com/browse/OSPRH-16632 , but I am not 100% sure. Please help us understand the path forward.

Expected behavior
A valid baremetal-deployment.yaml is generated after node is moved to Rhel9 role and inventory generation is not broken.

Bug impact
FFU process is blocked

Known workaround
None