You don't need to invent inflections. AFAIC, there's no reason to treat hortus different from virus (though I would take most people's word for the plural horti, but I would question both viri and virii). You don't need to give attestations for all inflections you add, but if challenged, you should be able to find them, most of the time. --184.108.40.206 22:04, 8 November 2012 (UTC)
It also opens a whole new can of worms. Imagine that all forms of a noun were attested except for the nominative singular. How would we add such a noun? Our current practice, which is very unlikely to change, is to reconstruct the nominative and add the entry there anyway. I believe that if we already do this for the nominative singular, I see no reason why we can't do it for any other cases, especially for languages that remain in active use.
Adding it to the nominative singular isn't a problem, I think. If the language has a case which is used more often and/or which predicts more cases, we might use that case as the place where all info is dropped. If we wouldn't say the nominative singular isn't attested, we would be lying buy omission.
AFAIK "loge" doesn't exist (as subj. of "logen"), didn't exist (because "logen" wasn't used), and won't exist (because the subjunctive isn't used anymore). We should only add it, once it had 0.01% of the usage of planking when that sense was kept/moved. For me, it's like koningetje, which shouldn't have an incoming link from koning#Dutch either (unless attested, of course). --220.127.116.11 01:48, 10 November 2012 (UTC)