The frontier generates too many messages if an IP address couldn't be determined

### Problem

The frontier logs
```
{"log":"2017-10-02 22:23:26,316 [pool-8-thread-57] ERROR [o.a.s.s.f.i.FrontierImpl ] - \u003cCould not recognize IP for http://hu.dbpedia.org/resource/Yale_Egyetem, unknown host\u003e\n","stream":"stdout","time":"2017-10-02T22:23:26.317066195Z"}
{"log":"2017-10-02 22:23:26,316 [pool-8-thread-57] ERROR [o.a.s.s.f.i.FrontierImpl ] - \u003cCouldn't determine the Inet address of \"http://hu.dbpedia.org/resource/Yale_Egyetem\". It will be ignored.\u003e\n","stream":"stdout","time":"2017-10-02T22:23:26.317089845Z"}
```
One of them with the log level `INFO` would be sufficient.

Additionally, the info might be cached. Otherwise the frontier will try to detremine the IP again after a short time.
```
{"log":"2017-10-02 22:34:46,168 [pool-8-thread-1] ERROR [o.a.s.s.f.i.FrontierImpl ] - \u003cCould not recognize IP for http://hu.dbpedia.org/resource/Anglia, unknown host\u003e\n","stream":"stdout","time":"2017-10-02T22:34:46.16824339Z"}
{"log":"2017-10-02 22:34:46,168 [pool-8-thread-1] ERROR [o.a.s.s.f.i.FrontierImpl ] - \u003cCouldn't determine the Inet address of \"http://hu.dbpedia.org/resource/Anglia\". It will be ignored.\u003e\n","stream":"stdout","time":"2017-10-02T22:34:46.16827295Z"}
```

Start by locating FrontierImpl, the component named in the reported log output, and trace how an unresolved host produces both messages and is retried. Compare the duplicate log paths and failed lookup handling; done means one sufficient info-level message is emitted and the failed address information is reused as requested.

Written by the indexing model from the issue text.