quoting Travis B.:

From reading that over, I would guess that <th> and <c> could also be /θ/ and /tʃ/, especially since the creators of Dorfish are most likely non-linguistically-aware English-speakers, and /θ/ and /tʃ/ are naturally sounds that a non-linguistically-aware English-speaker would put in a conlang.

This is the kind of logic I'm trying to avoid.  It's pretty obvious how the DF "languages" were put together (word generator set of roots, basic preset noun phrase forms based in English word order, not much else) if you look at them from an external perspective.  OTOH I'm happy to consider the possibility (indeed, probability) that transcriptions are flawed and/or by linguistically untrained native English speakers  

Something I forgot to look into was homonyms.  Initial results are interesting: they appear to exist, but some are so closely related they may have been misidentified as separate words.  Example: am, four different entries: open/yawn/speak/baby.  I'd say the first three have a clear semantic relation.

I'm also interested in the high proportion of vowel-initial words.  As Zontas suggests consonant loss is a good possibility (for example I suspect some sort of p > f > h > ∅ chain happened, which would explain the lack of /p/ and low frequency of /f/, but I wouldn't like to speculate on other lost consonants until I look into word structure more closely).

Still not sure how I'm going to extract a decent-sized corpus to study (what little) Dorfish grammar (there is). 

Edit:  As fuel for Zontas' paranoia, Dorfish does indeed include the word kikes.