I've added some commentary on the function on one of the blogs you've mentioned above, so please consult that as well.
But I stress - functions that are not documented are precisely that. *not documented*.
We do this for a reason - to ensure that functions do not get exposed before they are ready for production use. It's no use hounding me here at AskTOM for more information. In this particular case, whilst we've delivered full support for breed, age derivation and NLS handling, you can tell that function is not fully ready if you try pass one of those ridiculous new mixed-breeds into it - you'll get into rough territory.
SQL> select to_dog_year(date'2005-04-02','','NLS_BREED=Labradoodle') from dual;
select to_dog_year(date'2005-04-02','','NLS_BREED=Labradoodle') from dual;
ERROR at line 1:
ORA-00600: internal error code, arguments: [kcbmutt], , , , , , , , , , , 
Your hypotheses on OWN_SMOKE is *possibly* correct, but obviously, I cannot comment officially on that until we do fully document the function. Rest assured, by OOW 18 the functions will be public, thus validating your presentation submission on the topic. But also, as I commented on other blogs, as an Ace Director your are bound by NDA - if you publish information provided under the agreement, things might come back and bite you.
Once you see it in the official docs, that is how you'll know its fully supported.