I’d expect the behavior is correct now. There shouldn't be empty nodes. As some of them uses “default values” and system can see it as partly configured,
In any case the check is here https://github.com/vyos/vyos-1x/blob/ec18cc393591052fd1f021c4a62220ab2e537a2e/src/conf_mode/service_ntp.py#L68

And you’ll get the same behavior for every config.

With T3008 we moved from ntpd to chrony. This came with a restructuring of the CLI (mainly moving ntp out of system to services).

In addition the definition of a server was made mandatory.

The bug itself manifests at a more crucial point - config migration

Jan 23 20:51:18 LR3.wue3 vyos-router[1265]: Migration script error: /opt/vyatta/etc/config-migrate/migrate/ntp/1-to-2: [Errno 1] failed to run command: ['/opt/vyatta/etc/config-migrate/migrate/ntp/1-to-2', '/opt/vyatta/etc/config/config.boot']
Jan 23 20:51:18 LR3.wue3 vyos-router[1265]: returned: - op: copy old_path: ['system', 'ntp'] new_path: ['service', 'ntp']
Jan 23 20:51:18 LR3.wue3 vyos-router[1265]: - op: delete path: ['system', 'ntp']

A valid configuration for VyOS 1.3 is:

system {
  ntp {
  }
}

But VyOS 1.4 does not allow something like:

service {
  ntp {
  }
}

hence the configuration needs to be properly migrated "away" to not have an empty service ntp node

https://github.com/vyos/vyos-1x/pull/4315