Node03's disk was likely cloned from Node02's image (or a shared golden image) without running sysprep /generalize first. The rename to "azlocal03" was applied to the running OS but never written into the recovery/unattend partition. A reboot triggered a recovery-style boot pass that reapplied the original baked-in name ("azlocal02"), causing the hostname collision and cluster failure. Arc's clone-detection then flagged it since both machines shared one onboarding identity, that warning isn't a bug, it's working as intended.
Here are possible fixes to try;
- Reboot Node03 a couple more times to confirm the name sticks. If it reverts again, the disk needs proper re-imaging (generalized), don't touch Arc yet
- On each node: azcmagent disconnect (use --force-local-only if the Azure resource is already orphaned).
- In the portal, delete both broken Machine
- Re-register through the Azure Local cluster registration flow (Register-AzStackHCI), not bare azcmagent connect.
- Verify Name/Status/Manufacturer match the correct physical hosts.
Also check for a duplicate machine SID between the two nodes, since it's a common side effect of the same cloning issue.