It depends on how 'idempotent' the service calls are. If it's safe to retry the entire register method over and over again then it's pretty simple to keep track

It depends on how 'idempotent' the service calls are. If it's safe to retry the entire register method over and over again then it's pretty simple to keep track if anything blew up during the process and then just try from the top again. If those services cannot be safely retried then you'll need to deal with calling other operations on the services to roll back the changes if anything goes wrong. I worked on a system that had a similar problem (not idempotent and we actually didn't have way to delete/rollback changes through their service) so we had to log every call made for context. That way if any service failed we could retry the whole operation and skip over the things that were successful.