I have wondered about this, but even if botulinum or listeria are present, won't they grow very slowly in the refrigerator? I would think that defrosting for 12 hours or so at refrigerator temperatures would still be fairly safe. Or is the quantity that can be dangerous so minute that it doesn't matter?
Yes, defrosting vacuum sealed fish _in the fridge_ is safe, also in the package, for the next day as well, but not beyond. Defrosting on the countertop (or on the heating) is not, for the reasons GP stated.
I cold smoke meat so I need to be especially careful about this. The good news is, even though the spores are very hardy and can survive boiling at sea level atmospheric pressure, the botulism toxin (i.e. the chemical produced by the germs that does the actual damage) is easy to destroy by cooking over 85 degrees C.
Not inside meat, no. I don't think it's a problem for run of the mill hot-smoking or BBQ. For cold-smoked stuff I don't dare eat it raw yet but use the meat in stews. Cold smoked pork ribs add a 3rd dimension to thick soups like pasta fagioli or split pea soup, while sausages are great fried. I don't know how to safely attempt a prosciutto or lox though :-)
cured lox is easy, it's just a salt and sugar brine (cold smoked is something entirely different). If the salinity and sugar are high enough basically nothing can grow. Salt on its own, sugar on its own don't worry very well.
Perhaps its one of those things where “95% of the time you’ll be fine, but be careful if you do it 20 times, because botulism only needs to happen once.”
Probably just a locution, but be careful with that idea about numbers:
The probability of remaining in a safe zone of 95% for 20 consecutive times is ~36%. It ( (1-(1/n))^n ) approximates to 1/e.
The number of tries to reduce that 95% to a 50% coin toss is 13. It is not fully intuitive. One will put his threshold wherever one may think appropriate, but. A chance of tenth, a hundredth, a thousandth etc. for ten, a hundred, thousand etc. times approximates to 63.2%.
...Maybe a locution like 'Maybe 9 out of 10 times you'll be fine, but be careful about doing it 7 times' could work well (as 6.931 approximates the risk to a coin toss).
NB: This failure of statistical logic also shows up in "100 year storm" statistics.
The actual measurement is "storm with a 1% likelihood of occurring in any one year". The odds of that storm occurring in a century is actually about 63%, and of occurring in any ten year period, about 10%.
Add to that the fact that the measurement is based on a storm of a given magnitude (usually total rainfall / precipitation, wind speed, storm surge, etc.) occurring, and small changes in the likelihood of such events can dramatically change the rate at which a storm of a given magnitude is observed. If you live in/near flood-prone areas, keep an eye on, e.g., changes in what are considered flood stages (say, as flood control structures are added or removed), or to likelihoods of events of a given precipitation level or stream height.
The fact that most people don't know to remove the plastic and yet there are virtually no reported cases of people being poisoned by wrapped fish suggests that the danger is remote. It probably helps that if you fish get to the point where the bactera become active again it tends to smell bad and discolour.
