Over on Phonoloblog, Bruce Hayes has posted an interesting query about patterns of final voicing among Russian speakers. His data comes from Russian-Americans, native speakers now bilingual in English, and they seem to sometimes produce final voiced obstruents, contrary to expectation, so that the word for 'city' comes out as [’gorəd], rather than with a final [t].
I can't speak to all of his issues, but here are a couple of relevant points:
First, Tom Purnell and some of us (see the Journal of English Linguistics, 2005, along with other papers) have observed what is probably a related pattern among Wisconsin English spoken by people who grew up in what were then heavily German-speaking communities: The phonetic literature reports little (or no) actual glottal pulsing in coda obstruents for 'American English', but we found that these folks used voicing as the primary cue. Overall, they had no difference in vowel duration (longer before 'voiced'), which is usually thought to be a (or the) major cue — that's been claimed to be a universal pattern even. I don't know about their German, but they probably did this in English for sociolinguistic reasons: They were working to overcome one of the most salient aspects of a German accent. It's not hard to imagine that an earlier generation could have produced Bad with [d] under those kind of circumstances — at least those who got good enough in English to produce the distinction. (Remarkably, the youngest generation today looks like they are coming around to something that sounds like final devoicing in English.)
Second, Polish is reported to have the same basic pattern as Russian with regard to final devoicing. But the 1997 University of Wisconsin dissertation by Božena Tieszen, Final stop devoicing in Polish: An acoustic and historical account for incomplete neutralization, shows tremendous variation across Polish dialects.
Update, Sunday noon (by Mr. Verb): If you're interested in this topic, you should definitely look at the followup comments on that Phonoloblog post.