In on-prem while having Zync creating all provider and API routes, when new proxy domains are set containing more than 63 characters, Zync succeeds to create the corresponding routes in Openshift but they are rejected by the cluster router.

Once the invalid routes exist in Openshift, updating the domains in System has no repercussion in the cluster router anymore. The new domains will appear in System, but the old invalid ones remain in Openshift without new ones being added. The invalid routes need first to be manually deleted in Openshift and then force an update of the proxy domains in System.

For System, it may be hard to identify the way the routes are being created and maintained by the customer and therefore probably not a good idea to simply enforce the 63-character limit on the domains it generates. Perhaps Zync could acknowledge the failure and report back to System.

Dev note
We should limit the entire domain to 255 characters (which is done already), and also validate each part ("label") so it's no longer than 63 characters.
This is what we estimated. If there is a change, we re-estimate.

See this comment
and this one