Currently to avoid receiving resources and ops for parts of the domain model which do not work on an older slave we use <ignored-resources> in the slave host.xml to ignore them.

Instead we should only pull in parts of the domain model that is actually required on the slave for its servers to start. Once started we need to be able to update the slave with domain information as the config changes

The main things which can change affecting the data required are:
1) A domain's server-group, where the profile (and thus extensions) and the socket-binding group can be changed.
2) A host's server-config, where the server-group and the socket-binding-group override might be changed.

For 1) we can piggy-back missing data to the slave.

For 2) the slave needs to connect back to the DC, attempt to get the DC lock and pull down the required data. If the lock cannot be obtained, the operation should fail fast.